[Soil organic carbon mineralization of Black Locust forest in the deep soil layer of the hilly region of the Loess Plateau, China].

Science.gov (United States)

Ma, Xin-Xin; Xu, Ming-Xiang; Yang, Kai

2012-11-01

The deep soil layer (below 100 cm) stores considerable soil organic carbon (SOC). We can reveal its stability and provide the basis for certification of the deep soil carbon sinks by studying the SOC mineralization in the deep soil layer. With the shallow soil layer (0-100 cm) as control, the SOC mineralization under the condition (temperature 15 degrees C, the soil water content 8%) of Black Locust forest in the deep soil layer (100-400 cm) of the hilly region of the Loess Plateau was studied. The results showed that: (1) There was a downward trend in the total SOC mineralization with the increase of soil depth. The total SOC mineralization in the sub-deep soil (100-200 cm) and deep soil (200-400 cm) were equivalent to approximately 88.1% and 67.8% of that in the shallow layer (0-100 cm). (2) Throughout the carbon mineralization process, the same as the shallow soil, the sub-deep and deep soil can be divided into 3 stages. In the rapid decomposition phase, the ratio of the mineralization or organic carbon to the total mineralization in the sub-deep and deep layer (0-10 d) was approximately 50% of that in the shallow layer (0-17 d). In the slow decomposition phase, the ratio of organic carbon mineralization to total mineralization in the sub-deep, deep layer (11-45 d) was 150% of that in the shallow layer (18-45 d). There was no significant difference in this ratio among these three layers (46-62 d) in the relatively stable stage. (3) There was no significant difference (P > 0.05) in the mineralization rate of SOC among the shallow, sub-deep, deep layers. The stability of SOC in the deep soil layer (100-400 cm) was similar to that in the shallow soil layer and the SOC in the deep soil layer was also involved in the global carbon cycle. The change of SOC in the deep soil layer should be taken into account when estimating the effects of soil carbon sequestration in the Hilly Region of the Loess Plateau, China.

Beyond Retinal Layers: A Deep Voting Model for Automated Geographic Atrophy Segmentation in SD-OCT Images.

Science.gov (United States)

Ji, Zexuan; Chen, Qiang; Niu, Sijie; Leng, Theodore; Rubin, Daniel L

2018-01-01

To automatically and accurately segment geographic atrophy (GA) in spectral-domain optical coherence tomography (SD-OCT) images by constructing a voting system with deep neural networks without the use of retinal layer segmentation. An automatic GA segmentation method for SD-OCT images based on the deep network was constructed. The structure of the deep network was composed of five layers, including one input layer, three hidden layers, and one output layer. During the training phase, the labeled A-scans with 1024 features were directly fed into the network as the input layer to obtain the deep representations. Then a soft-max classifier was trained to determine the label of each individual pixel. Finally, a voting decision strategy was used to refine the segmentation results among 10 trained models. Two image data sets with GA were used to evaluate the model. For the first dataset, our algorithm obtained a mean overlap ratio (OR) 86.94% ± 8.75%, absolute area difference (AAD) 11.49% ± 11.50%, and correlation coefficients (CC) 0.9857; for the second dataset, the mean OR, AAD, and CC of the proposed method were 81.66% ± 10.93%, 8.30% ± 9.09%, and 0.9952, respectively. The proposed algorithm was capable of improving over 5% and 10% segmentation accuracy, respectively, when compared with several state-of-the-art algorithms on two data sets. Without retinal layer segmentation, the proposed algorithm could produce higher segmentation accuracy and was more stable when compared with state-of-the-art methods that relied on retinal layer segmentation results. Our model may provide reliable GA segmentations from SD-OCT images and be useful in the clinical diagnosis of advanced nonexudative AMD. Based on the deep neural networks, this study presents an accurate GA segmentation method for SD-OCT images without using any retinal layer segmentation results, and may contribute to improved understanding of advanced nonexudative AMD.

Quantitative evaluation of deep and shallow tissue layers' contribution to fNIRS signal using multi-distance optodes and independent component analysis.

Science.gov (United States)

Funane, Tsukasa; Atsumori, Hirokazu; Katura, Takusige; Obata, Akiko N; Sato, Hiroki; Tanikawa, Yukari; Okada, Eiji; Kiguchi, Masashi

2014-01-15

To quantify the effect of absorption changes in the deep tissue (cerebral) and shallow tissue (scalp, skin) layers on functional near-infrared spectroscopy (fNIRS) signals, a method using multi-distance (MD) optodes and independent component analysis (ICA), referred to as the MD-ICA method, is proposed. In previous studies, when the signal from the shallow tissue layer (shallow signal) needs to be eliminated, it was often assumed that the shallow signal had no correlation with the signal from the deep tissue layer (deep signal). In this study, no relationship between the waveforms of deep and shallow signals is assumed, and instead, it is assumed that both signals are linear combinations of multiple signal sources, which allows the inclusion of a "shared component" (such as systemic signals) that is contained in both layers. The method also assumes that the partial optical path length of the shallow layer does not change, whereas that of the deep layer linearly increases along with the increase of the source-detector (S-D) distance. Deep- and shallow-layer contribution ratios of each independent component (IC) are calculated using the dependence of the weight of each IC on the S-D distance. Reconstruction of deep- and shallow-layer signals are performed by the sum of ICs weighted by the deep and shallow contribution ratio. Experimental validation of the principle of this technique was conducted using a dynamic phantom with two absorbing layers. Results showed that our method is effective for evaluating deep-layer contributions even if there are high correlations between deep and shallow signals. Next, we applied the method to fNIRS signals obtained on a human head with 5-, 15-, and 30-mm S-D distances during a verbal fluency task, a verbal working memory task (prefrontal area), a finger tapping task (motor area), and a tetrametric visual checker-board task (occipital area) and then estimated the deep-layer contribution ratio. To evaluate the signal separation

Spatial extent and dissipation of the deep chlorophyll layer in Lake Ontario during the Lake Ontario lower foodweb assessment, 2003 and 2008

Science.gov (United States)

Watkins, J. M.; Weidel, Brian M.; Rudstam, L. G.; Holek, K. T.

2014-01-01

Increasing water clarity in Lake Ontario has led to a vertical redistribution of phytoplankton and an increased importance of the deep chlorophyll layer in overall primary productivity. We used in situ fluorometer profiles collected in lakewide surveys of Lake Ontario in 2008 to assess the spatial extent and intensity of the deep chlorophyll layer. In situ fluorometer data were corrected with extracted chlorophyll data using paired samples from Lake Ontario collected in August 2008. The deep chlorophyll layer was present offshore during the stratified conditions of late July 2008 with maximum values from 4-13 μg l-1 corrected chlorophyll a at 10 to 17 m depth within the metalimnion. Deep chlorophyll layer was closely associated with the base of the thermocline and a subsurface maximum of dissolved oxygen, indicating the feature's importance as a growth and productivity maximum. Crucial to the deep chlorophyll layer formation, the photic zone extended deeper than the surface mixed layer in mid-summer. The layer extended through most of the offshore in July 2008, but was not present in the easternmost transect that had a deeper surface mixed layer. By early September 2008, the lakewide deep chlorophyll layer had dissipated. A similar formation and dissipation was observed in the lakewide survey of Lake Ontario in 2003.

Evaluation of vertical coordinate and vertical mixing algorithms in the HYbrid-Coordinate Ocean Model (HYCOM)

Science.gov (United States)

Halliwell, George R.

Vertical coordinate and vertical mixing algorithms included in the HYbrid Coordinate Ocean Model (HYCOM) are evaluated in low-resolution climatological simulations of the Atlantic Ocean. The hybrid vertical coordinates are isopycnic in the deep ocean interior, but smoothly transition to level (pressure) coordinates near the ocean surface, to sigma coordinates in shallow water regions, and back again to level coordinates in very shallow water. By comparing simulations to climatology, the best model performance is realized using hybrid coordinates in conjunction with one of the three available differential vertical mixing models: the nonlocal K-Profile Parameterization, the NASA GISS level 2 turbulence closure, and the Mellor-Yamada level 2.5 turbulence closure. Good performance is also achieved using the quasi-slab Price-Weller-Pinkel dynamical instability model. Differences among these simulations are too small relative to other errors and biases to identify the "best" vertical mixing model for low-resolution climate simulations. Model performance deteriorates slightly when the Kraus-Turner slab mixed layer model is used with hybrid coordinates. This deterioration is smallest when solar radiation penetrates beneath the mixed layer and when shear instability mixing is included. A simulation performed using isopycnic coordinates to emulate the Miami Isopycnic Coordinate Ocean Model (MICOM), which uses Kraus-Turner mixing without penetrating shortwave radiation and shear instability mixing, demonstrates that the advantages of switching from isopycnic to hybrid coordinates and including more sophisticated turbulence closures outweigh the negative numerical effects of maintaining hybrid vertical coordinates.

Deep-Layer Microvasculature Dropout by Optical Coherence Tomography Angiography and Microstructure of Parapapillary Atrophy.

Science.gov (United States)

Suh, Min Hee; Zangwill, Linda M; Manalastas, Patricia Isabel C; Belghith, Akram; Yarmohammadi, Adeleh; Akagi, Tadamichi; Diniz-Filho, Alberto; Saunders, Luke; Weinreb, Robert N

2018-04-01

To investigate the association between the microstructure of β-zone parapapillary atrophy (βPPA) and parapapillary deep-layer microvasculature dropout assessed by optical coherence tomography angiography (OCT-A). Thirty-seven eyes with βPPA devoid of the Bruch's membrane (BM) (γPPA) ranging between completely absent and discontinuous BM were matched by severity of the visual field (VF) damage with 37 eyes with fully intact BM (βPPA+BM) based on the spectral-domain (SD) OCT imaging. Parapapillary deep-layer microvasculature dropout was defined as a dropout of the microvasculature within choroid or scleral flange in the βPPA on the OCT-A. The widths of βPPA, γPPA, and βPPA+BM were measured on six radial SD-OCT images. Prevalence of the dropout was compared between eyes with and without γPPA. Logistic regression was performed for evaluating association of the dropout with the width of βPPA, γPPA, and βPPA+BM, and the γPPA presence. Eyes with γPPA had significantly higher prevalence of the dropout than did those without γPPA (75.7% versus 40.8%; P = 0.004). In logistic regression, presence and longer width of the γPPA, worse VF mean deviation, and presence of focal lamina cribrosa defects were significantly associated with the dropout (P 0.10). Parapapillary deep-layer microvasculature dropout was associated with the presence and larger width of γPPA, but not with the βPPA+BM width. Presence and width of the exposed scleral flange, rather than the retinal pigmented epithelium atrophy, may be associated with deep-layer microvasculature dropout.

Autonomous Gliders Observed Physical and Biogeochemical Interplay at Submesoscale during Deep Convection in the Gulf of Lions (NW Mediterranean)

Science.gov (United States)

Bosse, A.; Testor, P.; Damien, P.; D'Ortenzio, F.; Prieur, L. M.; Estournel, C.; Marsaleix, P.; Mortier, L.

2016-02-01

Since 2010, sustained observations of the circulation and water properties of the NW Mediterranean Sea have been carried out by gliders in the framework of the MOOSE observatory (Mediterranean Ocean Observatory System for the Environment: http://www.moose-network.fr/). They regularly sampled the wintertime Northern Current (NC), the deep convection zone as well as the North Balearic Front (NBF) collecting a great amount of physical and biogeochemical measurements.During periods of deep convection, the offshore mixed layer can reach great depths (>2300 m) in the Gulf of Lions. Baroclinic fronts then become very intense and reveal a lot of variability at submesoscale in the upper 500 m or so. In terms of process, symmetric instability has been evidenced to occurr during strong wind events by gliders measurements. Complementary analysis performed with the help of a high-resolution regional model (dx,dy=1 km) highlight the prominent role of downfront winds in triggering this instability. Important vertical exchanges of oceanic tracers at the front approximately aligned with isopycnals of magnitude O(100m/day) occur in response to this strong atmospheric forcing. Finally, gliders measurements of Chl-a fluorescence show how this frontal instability seems to stimulate phytoplankton growth in frontal regions during harsh wintertime conditions.

Reconstruction of Hyaline Cartilage Deep Layer Properties in 3-Dimensional Cultures of Human Articular Chondrocytes.

Science.gov (United States)

Nanduri, Vibudha; Tattikota, Surendra Mohan; T, Avinash Raj; Sriramagiri, Vijaya Rama Rao; Kantipudi, Suma; Pande, Gopal

2014-06-01

Articular cartilage (AC) injuries and malformations are commonly noticed because of trauma or age-related degeneration. Many methods have been adopted for replacing or repairing the damaged tissue. Currently available AC repair methods, in several cases, fail to yield good-quality long-lasting results, perhaps because the reconstructed tissue lacks the cellular and matrix properties seen in hyaline cartilage (HC). To reconstruct HC tissue from 2-dimensional (2D) and 3-dimensional (3D) cultures of AC-derived human chondrocytes that would specifically exhibit the cellular and biochemical properties of the deep layer of HC. Descriptive laboratory study. Two-dimensional cultures of human AC-derived chondrocytes were established in classical medium (CM) and newly defined medium (NDM) and maintained for a period of 6 weeks. These cells were suspended in 2 mm-thick collagen I gels, placed in 24-well culture inserts, and further cultured up to 30 days. Properties of chondrocytes, grown in 2D cultures and the reconstructed 3D cartilage tissue, were studied by optical and scanning electron microscopic techniques, immunohistochemistry, and cartilage-specific gene expression profiling by reverse transcription polymerase chain reaction and were compared with those of the deep layer of native human AC. Two-dimensional chondrocyte cultures grown in NDM, in comparison with those grown in CM, showed more chondrocyte-specific gene activity and matrix properties. The NDM-grown chondrocytes in 3D cultures also showed better reproduction of deep layer properties of HC, as confirmed by microscopic and gene expression analysis. The method used in this study can yield cartilage tissue up to approximately 1.6 cm in diameter and 2 mm in thickness that satisfies the very low cell density and matrix composition properties present in the deep layer of normal HC. This study presents a novel and reproducible method for long-term culture of AC-derived chondrocytes and reconstruction of cartilage

Flow-dependent assimilation of sea surface temperature in isopycnal coordinates with the Norwegian Climate Prediction Model

Directory of Open Access Journals (Sweden)

François Counillon

2016-12-01

Full Text Available We document a pilot stochastic re-analysis computed by assimilating sea surface temperature (SST anomalies into the ocean component of the coupled Norwegian Climate Prediction Model (NorCPM for the period 1950–2010 (doi: 10.11582/2016.00002. NorCPM is based on the Norwegian Earth System Model and uses the ensemble Kalman filter for data assimilation (DA. Here, we assimilate SST from the stochastic HadISST2 historical reconstruction. The accuracy, reliability and drift are investigated using both assimilated and independent observations. NorCPM is slightly overdispersive against assimilated observations but shows stable performance through the analysis period. It demonstrates skills against independent measurements: sea surface height, heat and salt content, in particular in the Equatorial and North Pacific, the North Atlantic Subpolar Gyre (SPG region and the Nordic Seas. Furthermore, NorCPM provides a reliable monitoring of the SPG index and represents the vertical temperature variability there, in good agreement with observations. The monitoring of the Atlantic meridional overturning circulation is also encouraging. The benefit of using a flow-dependent assimilation method and constructing the covariance in isopycnal coordinates are investigated in the SPG region. Isopycnal coordinates discretisation is found to better capture the vertical structure than standard depth-coordinate discretisation, because it leads to a deeper influence of the assimilated surface observations. The vertical covariance shows a pronounced seasonal and decadal variability that highlights the benefit of flow-dependent DA method. This study demonstrates the potential of NorCPM to compute an ocean re-analysis for the 19th and 20th centuries when SST observations are available.

Reactive Fe(II) layers in deep-sea sediments

Science.gov (United States)

König, Iris; Haeckel, Matthias; Drodt, Matthias; Suess, Erwin; Trautwein, Alfred X.

1999-05-01

The percentage of the structural Fe(II) in clay minerals that is readily oxidized to Fe(III) upon contact with atmospheric oxygen was determined across the downcore tan-green color change in Peru Basin sediments. This latent fraction of reactive Fe(II) was only found in the green strata, where it proved to be large enough to constitute a deep reaction layer with respect to the pore water O 2 and NO 3-. Large variations were detected in the proportion of the reactive Fe(II) concentration to the organic matter content along core profiles. Hence, the commonly observed tan-green color change in marine sediments marks the top of a reactive Fe(II) layer, which may represent the major barrier to the movement of oxidation fronts in pelagic subsurface sediments. This is also demonstrated by numerical model simulations. The findings imply that geochemical barriers to pore water oxidation fronts form diagenetically in the sea floor wherever the stage of iron reduction is reached, provided that the sediments contain a significant amount of structural iron in clay minerals.

Seasonal and inter-annual temperature variability in the bottom waters over the western Black Sea shelf

Directory of Open Access Journals (Sweden)

G. I. Shapiro

2011-09-01

Full Text Available Long-term changes in the state of the Bottom Shelf Water (BSW on the Western shelf of the Black Sea are assessed using analysis of intra-seasonal and inter-annual temperature variations. For the purpose of this study the BSW is defined as such shelf water mass between the seabed and the upper mixed layer (bounded by the σ_θ = 14.2 isopycnal which has limited ability to mix vertically with oxygen-rich surface waters during the warm season due to formation of a seasonal pycnocline. A long-term time series of temperature anomalies in the BSW is constructed from in-situ observations taken over the 2nd half of the 20th century. The BSW is shown to occupy nearly half of the shelf area during the summer stratification period (May–November.The results reveal a warm phase in the 1960s/70s, followed by a cold phase between 1985 and 1995 and a further warming after 1995. The transition between the warm and cold periods coincides with a regime shift in the Black Sea ecosystem. While it was confirmed that the memory of winter convection is well preserved over the following months in the deep sea, the signal of winter cooling in the BSW significantly reduces during the warm season. The potential of the BSW to ventilate horizontally during the warm season with the deep-sea waters is assessed using isopycnic analysis of temperature variations. It is shown that temperature in the BSW is stronger correlated with the temperature of Cold Intermediate Waters (CIW in the deep sea than with the severity of the previous winters, thus indicating that the isopycnal exchanges with the deep sea are more important for inter-annual/inter-decadal variability of the BSW on the western Black Sea shelf than effects of winter convection on the shelf itself.

Modelling deep convection and its impacts on the tropical tropopause layer

Directory of Open Access Journals (Sweden)

J. S. Hosking

2010-11-01

Full Text Available The UK Met Office's Unified Model is used at a climate resolution (N216, ~0.83°×~0.56°, ~60 km to assess the impact of deep tropical convection on the structure of the tropical tropopause layer (TTL. We focus on the potential for rapid transport of short-lived ozone depleting species to the stratosphere by rapid convective uplift. The modelled horizontal structure of organised convection is shown to match closely with signatures found in the OLR satellite data. In the model, deep convective elevators rapidly lift air from 4–5 km up to 12–14 km. The influx of tropospheric air entering the TTL (11–12 km is similar for all tropical regions with most convection stopping below ~14 km. The tropical tropopause is coldest and driest between November and February, coinciding with the greatest upwelling over the tropical warm pool. As this deep convection is co-located with bromine-rich biogenic coastal emissions, this period and location could potentially be the preferential gateway for stratospheric bromine.

Physical oceanographic characteristics influencing the dispersion of dissolved tracers released at the sea floor in selected deep ocean study areas

International Nuclear Information System (INIS)

Kupferman, S.L.; Moore, D.E.

1981-02-01

Scenarios which follow the development in space and time of the concentration field of a dissolved tracer released at the sea floor are presented for a Pacific and two Atlantic study areas. The scenarios are closely tied to available data by means of simple analytical models and proceed in stages from short time and space scales in the immediate vicinity of a release point to those scales characteristic of ocean basins. The concepts of internal mixing time and residence time in the benthic mixed layer, useful for developing an intuitive feeling for the behavior of a tracer in this feature, are introduced and discussed. We also introduce the concept of domain of occupation, which is useful in drawing distinctions between mixing and stirring in the ocean. From this study it is apparent that reliable estimation of mixing will require careful consideration of the dynamics of the eddy fields in the ocean. Another area in which more information is urgently needed is in the relation of deep isopycnal structure and bottom topography to local near-bottom circulation

Modification of the deep salinity-maximum in the Southern Ocean by circulation in the Antarctic Circumpolar Current and the Weddell Gyre

Science.gov (United States)

Donnelly, Matthew; Leach, Harry; Strass, Volker

2017-07-01

The evolution of the deep salinity-maximum associated with the Lower Circumpolar Deep Water (LCDW) is assessed using a set of 37 hydrographic sections collected over a 20-year period in the Southern Ocean as part of the WOCE/CLIVAR programme. A circumpolar decrease in the value of the salinity-maximum is observed eastwards from the North Atlantic Deep Water (NADW) in the Atlantic sector of the Southern Ocean through the Indian and Pacific sectors to Drake Passage. Isopycnal mixing processes are limited by circumpolar fronts, and in the Atlantic sector, this acts to limit the direct poleward propagation of the salinity signal. Limited entrainment occurs into the Weddell Gyre, with LCDW entering primarily through the eddy-dominated eastern limb. A vertical mixing coefficient, κV of (2.86 ± 1.06) × 10-4 m2 s-1 and an isopycnal mixing coefficient, κI of (8.97 ± 1.67) × 102 m2 s-1 are calculated for the eastern Indian and Pacific sectors of the Antarctic Circumpolar Current (ACC). A κV of (2.39 ± 2.83) × 10-5 m2 s-1, an order of magnitude smaller, and a κI of (2.47 ± 0.63) × 102 m2 s-1, three times smaller, are calculated for the southern and eastern Weddell Gyre reflecting a more turbulent regime in the ACC and a less turbulent regime in the Weddell Gyre. In agreement with other studies, we conclude that the ACC acts as a barrier to direct meridional transport and mixing in the Atlantic sector evidenced by the eastward propagation of the deep salinity-maximum signal, insulating the Weddell Gyre from short-term changes in NADW characteristics.

Penaeoid and sergestoid shrimps from the deep scattering layer (DSL) in the Arabian Sea

Digital Repository Service at National Institute of Oceanography (India)

Karuppasamy, P.K.; Menon, N.G.

Results of a preliminary study on the occurrence and distribution of seventeen species of Penaeoid and Sergestoid shrimps from the deep scattering layer (DSL) of the Indian EEZ of Arabian Sea are presented here based on the IKMT samples collected...

Chlorophyll modulation of mixed layer thermodynamics in a mixed-layer isopycnal general circulation model - An example from Arabian Sea and Equatorial Pacific

Digital Repository Service at National Institute of Oceanography (India)

Nakamoto, S.; PrasannaKumar, S.; Oberhuber, J.M.; Saito, H.; Muneyama, K.

and supported by quasi-steady upwelling. Remotely sensed chlorophyll pigment concentrations from the Coastal Zone Color Scanner (CZCS) are used to investigate the chlorophyll modulation of ocean mixed layer thermodynamics in a bulk mixed-layer model, embedded...

Towards a more consistent picture of isopycnal mixing in climate models

Science.gov (United States)

Gnanadesikan, A.; Pradal, M. A. S.; Koszalka, I.; Abernathey, R. P.

2014-12-01

The stirring of tracers by mesoscale eddies along isopycnal surfaces is often represented in coarse-resolution models by the Redi diffusion parameter ARedi. Theoretical treatments of ARedi often assume it should scale as the eddy energy or the growth rate of mesoscale eddies,. producing a picture where it is high in boundary currents and low )of order a few hundred m2/s) in the gyre interiors. However, observational estimates suggest that ARedi should be very large (of order thousands of m2/s) in the gyre interior. We present results of recent simulations comparing a range of spatially constant values ARedi (with values of 400, 800, 1200 and 2400 m2/s) to a spatially resolved estimate based on altimetry and a zonally averaged version of the same estimate. In general, increasing the ARedi coefficient destratifies and warms the high latitudes. Relative to our control simulation, the spatially dependent coefficient is lower in the Southern Ocean, but high in the North Pacific, and so the temperature changes mirror this. We also examine the response of ocean hypoxia to these changes. In general, the zonally averaged version of the altimetry-based estimate of ARedi does not capture the full 2d representation.

An Algorithm to Generate Deep-Layer Temperatures from Microwave Satellite Observations for the Purpose of Monitoring Climate Change. Revised

Science.gov (United States)

Goldberg, Mitchell D.; Fleming, Henry E.

1994-01-01

An algorithm for generating deep-layer mean temperatures from satellite-observed microwave observations is presented. Unlike traditional temperature retrieval methods, this algorithm does not require a first guess temperature of the ambient atmosphere. By eliminating the first guess a potentially systematic source of error has been removed. The algorithm is expected to yield long-term records that are suitable for detecting small changes in climate. The atmospheric contribution to the deep-layer mean temperature is given by the averaging kernel. The algorithm computes the coefficients that will best approximate a desired averaging kernel from a linear combination of the satellite radiometer's weighting functions. The coefficients are then applied to the measurements to yield the deep-layer mean temperature. Three constraints were used in deriving the algorithm: (1) the sum of the coefficients must be one, (2) the noise of the product is minimized, and (3) the shape of the approximated averaging kernel is well-behaved. Note that a trade-off between constraints 2 and 3 is unavoidable. The algorithm can also be used to combine measurements from a future sensor (i.e., the 20-channel Advanced Microwave Sounding Unit (AMSU)) to yield the same averaging kernel as that based on an earlier sensor (i.e., the 4-channel Microwave Sounding Unit (MSU)). This will allow a time series of deep-layer mean temperatures based on MSU measurements to be continued with AMSU measurements. The AMSU is expected to replace the MSU in 1996.

The Deep Atmospheric Boundary Layer and Its Significance to the Stratosphere and Troposphere Exchange over the Tibetan Plateau

Science.gov (United States)

Chen, Xuelong; Añel, Juan A.; Su, Zhongbo; de la Torre, Laura; Kelder, Hennie; van Peet, Jacob; Ma, Yaoming

2013-01-01

In this study the depth of the atmospheric boundary layer (ABL) over the Tibetan Plateau was measured during a regional radiosonde observation campaign in 2008 and found to be deeper than indicated by previously measurements. Results indicate that during fair weather conditions on winter days, the top of the mixed layers can be up to 5 km above the ground (9.4 km above sea level). Measurements also show that the depth of the ABL is quite distinct for three different periods (winter, monsoon-onset, and monsoon seasons). Turbulence at the top of a deep mixing layer can rise up to the upper troposphere. As a consequence, as confirmed by trajectory analysis, interaction occurs between deep ABLs and the low tropopause during winter over the Tibetan Plateau. PMID:23451108

Layer dividing and zone dividing of physical property of crust and deep structure in Jiangxi province

International Nuclear Information System (INIS)

Li Chunhua; Yang Yaxin; Gong Yuling; Huang Linping

2001-01-01

On the base of summing experiences both at home and abroad, the Bugar gravitative anomalies are studied by major means of data processing. According to the anomalous character, three layer crust models (surface layer, middle layer in region and material layer under crust) are built up, depth of upper and bottom surfaces for every layer is calculated quantitatively, their varied characters of depth are studied and deep geological tectonics are outlined. The 'density' and 'mass' of every layer are calculated, and according to these two parameters, the shallow geological tectonics are researched. The relation-factor R between the surface altitude and Bugar gravitative anomalies are calculated and the stable or unstable crust zones are divided. The favorable mine zones for uranium deposit in Jiangxi Province are outlined

An environmental model study of the deep layers of the North East Atlantic

International Nuclear Information System (INIS)

Bork, I.

1989-01-01

The field work of the north Atlantic monitoring Program (NOAMP) was supplemented by numerical simulations of the transport of radionuclides in the North Atlantic Ocean by annual mean flows and mixing processes. During the last year of NOAMP, a different attempt was made to compute the current field and three-dimensional trajectories of particles released in deep layers of the NOAMP area. It is the subject of this paper. The model used is of Bryan/Semtner type, but with smoothed topography and climatological (winter) temperature and salinity data. The results form a compromise between interpretation of climatological temperature and salinity data and the complete prediction of the current field by prognostic calculations, which yields a deep flow pattern that agrees with some ideas of the abyssal circulation

Role of ocean isopycnal mixing in setting the uptake of anthropogenic carbon

Science.gov (United States)

Gnanadesikan, A.; Pradal, M. A. S.; Abernathey, R. P.

2014-12-01

The magnitude of the isopycnal stirring coefficient ARedi is poorly constrained from data and varies greatly across Earth System Models. This paper documents the impact of such uncertainty on the oceanic carbon cycle. We compare six spatial representations of ARedi. Four constant values (400, 800, 1200 and 2400 m2/s) are used to explore the difference between using the low values found in many models and the higher values seen in observational estimates. Models are also run with two spatially dependent values of ARedi based on altimetry, one which captures the fully two-dimensional structure of the mixing coefficient, the other of which looks at the zonally averaged structure alone. Under global warming significant changes are seen in the biological pump in convective regions, but these changes are largely locally compensated by changes in preformed DIC. Instead, differences in anthropogenic uptake of carbon are largely centered in the tropics, and can be well described in terms of a relatively simple diffusive approximation. Using ideal age as a tracer can give insight into the expected behavior of the models. The rate of oceanic mixing represents a quantitatively significant uncertainty in future projections of the global carbon cycle, amounting to about 20% of the oceanic uptake.

Jetting from impact of a spherical drop with a deep layer

Science.gov (United States)

Zhang, Li; Toole, Jameson; Fazzaa, Kamel; Deegan, Robert; Deegan Group Team; X-Ray Science Division, Advanced Photon Source Collaboration

2011-11-01

We performed an experimental study of jets during the impact of a spherical drop with a deep layer of same liquid. Using high speed optical and X-ray imaging, we observe two types of jets: the so-called ejecta sheet which emerges almost immediately after impact and the lamella which emerges later. For high Reynolds number the two jets are distinct, while for low Reynolds number the two jets combine into a single continuous jet. We also measured the emergence time, speed, and position of the ejecta sheet and found simple scaling relations for these quantities.

Deep Galaxy: Classification of Galaxies based on Deep Convolutional Neural Networks

OpenAIRE

Khalifa, Nour Eldeen M.; Taha, Mohamed Hamed N.; Hassanien, Aboul Ella; Selim, I. M.

2017-01-01

In this paper, a deep convolutional neural network architecture for galaxies classification is presented. The galaxy can be classified based on its features into main three categories Elliptical, Spiral, and Irregular. The proposed deep galaxies architecture consists of 8 layers, one main convolutional layer for features extraction with 96 filters, followed by two principles fully connected layers for classification. It is trained over 1356 images and achieved 97.272% in testing accuracy. A c...

A Fusion Face Recognition Approach Based on 7-Layer Deep Learning Neural Network

Directory of Open Access Journals (Sweden)

Jianzheng Liu

2016-01-01

Full Text Available This paper presents a method for recognizing human faces with facial expression. In the proposed approach, a motion history image (MHI is employed to get the features in an expressive face. The face can be seen as a kind of physiological characteristic of a human and the expressions are behavioral characteristics. We fused the 2D images of a face and MHIs which were generated from the same face’s image sequences with expression. Then the fusion features were used to feed a 7-layer deep learning neural network. The previous 6 layers of the whole network can be seen as an autoencoder network which can reduce the dimension of the fusion features. The last layer of the network can be seen as a softmax regression; we used it to get the identification decision. Experimental results demonstrated that our proposed method performs favorably against several state-of-the-art methods.

Light penetration structures the deep acoustic scattering layers in the global ocean

DEFF Research Database (Denmark)

Aksnes, Dag L.; Rostad, Anders; Kaartvedt, Stein

2017-01-01

The deep scattering layer (DSL) is a ubiquitous acoustic signature found across all oceans and arguably the dominant feature structuring the pelagic open ocean ecosystem. It is formed by mesopelagic fishes and pelagic invertebrates. The DSL animals are an important food source for marine megafauna...... distributions with hypoxic waters. In enhancing understanding of this phenomenon, our results should improve the ability to predict and model the dynamics of one of the largest animal biomass components on earth, with key roles in the oceanic biological carbon pump and food web....

Methane oxidation and methane fluxes in the ocean surface layer and deep anoxic waters

Science.gov (United States)

Ward, B. B.; Kilpatrick, K. A.; Novelli, P. C.; Scranton, M. I.

1987-01-01

Measured biological oxidation rates of methane in near-surface waters of the Cariaco Basin are compared with the diffusional fluxes computed from concentration gradients of methane in the surface layer. Methane fluxes and oxidation rates were investigated in surface waters, at the oxic/anoxic interface, and in deep anoxic waters. It is shown that the surface-waters oxidation of methane is a mechanism which modulates the flux of methane from marine waters to the atmosphere.

Handwritten Devanagari Character Recognition Using Layer-Wise Training of Deep Convolutional Neural Networks and Adaptive Gradient Methods

Directory of Open Access Journals (Sweden)

Mahesh Jangid

2018-02-01

Full Text Available Handwritten character recognition is currently getting the attention of researchers because of possible applications in assisting technology for blind and visually impaired users, human–robot interaction, automatic data entry for business documents, etc. In this work, we propose a technique to recognize handwritten Devanagari characters using deep convolutional neural networks (DCNN which are one of the recent techniques adopted from the deep learning community. We experimented the ISIDCHAR database provided by (Information Sharing Index ISI, Kolkata and V2DMDCHAR database with six different architectures of DCNN to evaluate the performance and also investigate the use of six recently developed adaptive gradient methods. A layer-wise technique of DCNN has been employed that helped to achieve the highest recognition accuracy and also get a faster convergence rate. The results of layer-wise-trained DCNN are favorable in comparison with those achieved by a shallow technique of handcrafted features and standard DCNN.

Mixed layer depth calculation in deep convection regions in ocean numerical models

Science.gov (United States)

Courtois, Peggy; Hu, Xianmin; Pennelly, Clark; Spence, Paul; Myers, Paul G.

2017-12-01

Mixed Layer Depths (MLDs) diagnosed by conventional numerical models are generally based on a density difference with the surface (e.g., 0.01 kg.m-3). However, the temperature-salinity compensation and the lack of vertical resolution contribute to over-estimated MLD, especially in regions of deep convection. In the present work, we examined the diagnostic MLD, associated with the deep convection of the Labrador Sea Water (LSW), calculated with a simple density difference criterion. The over-estimated MLD led us to develop a new tool, based on an observational approach, to recalculate MLD from model output. We used an eddy-permitting, 1/12° regional configuration of the Nucleus for European Modelling of the Ocean (NEMO) to test and discuss our newly defined MLD. We compared our new MLD with that from observations, and we showed a major improvement with our new algorithm. To show the new MLD is not dependent on a single model and its horizontal resolution, we extended our analysis to include 1/4° eddy-permitting simulations, and simulations using the Modular Ocean Model (MOM) model.

Property Changes of Abyssal Waters in the Western Tropical Atlantic

Science.gov (United States)

Herrford, Josefine; Brandt, Peter; Zenk, Walter

2017-04-01

Flowing northward towards the equator, Antarctic Bottom Water (AABW) encounters the lighter overlying North Atlantic Deep Water (NADW), both water masses creating an abyssal stratification and gradually mixing across their interface. Changes in the associated water mass formation and/or along-path transformation, observable in the evolution of water mass volume and characteristics, might impact the deep oceans uptake of anthropogenic CO2 or its contribution to global sea level rise. We compile historic and recent shipboard measurements of hydrography and velocity to provide a comprehensive view on water mass distribution, pathways, along-path transformation and long-term temperature changes of abyssal waters in the western South and Equatorial Atlantic. We are able to confirm previous results showing that the northwest corner of the Brazil Basin represents a splitting point for the southward/northward flow of NADW/AABW. The available measurements sample water mass transformation along the two major routes for deep and bottom waters in the tropical to South Atlantic - along the deep western boundary and eastward, parallel to the equator - as well as the hot spots of extensive mixing. We find lower NADW and lighter AABW to form a highly interactive transition layer in the northern Brazil Basin. The AABW north of 5°S is relatively homogeneous with only lighter AABW being able to pass through the Equatorial Channel (EQCH) into the North Atlantic. Spanning a period of 26 years, our data also allow an estimation of long-term temperature trends in abyssal waters. We find a warming of 2.5 ± 0.7•10-3 °C yr-1 of the waters in the northern Brazil Basin being colder than 0.6 °C throughout the period 1989-2014 and can relate that warming to a thinning of the dense AABW layer. While isopycnal heave is the dominant effect defining the vertical distribution of temperature trends on isobars, we also find temperature changes on isopycnals in the transition layer the lower NADW

Comparison of the ocean surface vector winds over the Nordic Seas and their application for ocean modeling

Science.gov (United States)

Dukhovskoy, Dmitry; Bourassa, Mark

2017-04-01

Ocean processes in the Nordic Seas and northern North Atlantic are strongly controlled by air-sea heat and momentum fluxes. The predominantly cyclonic, large-scale atmospheric circulation brings the deep ocean layer up to the surface preconditioning the convective sites in the Nordic Seas for deep convection. In winter, intensive cooling and possibly salt flux from newly formed sea ice erodes the near-surface stratification and the mixed layer merges with the deeper domed layer, exposing the very weakly stratified deep water mass to direct interaction with the atmosphere. Surface wind is one of the atmospheric parameters required for estimating momentum and turbulent heat fluxes to the sea ice and ocean surface. In the ocean models forced by atmospheric analysis, errors in surface wind fields result in errors in air-sea heat and momentum fluxes, water mass formation, ocean circulation, as well as volume and heat transport in the straits. The goal of the study is to assess discrepancies across the wind vector fields from reanalysis data sets and scatterometer-derived gridded products over the Nordic Seas and northern North Atlantic and to demonstrate possible implications of these differences for ocean modeling. The analyzed data sets include the reanalysis data from the National Center for Environmental Prediction Reanalysis 2 (NCEPR2), Climate Forecast System Reanalysis (CFSR), Arctic System Reanalysis (ASR) and satellite wind products Cross-Calibrated Multi-Platform (CCMP) wind product version 1.1 and recently released version 2.0, and Remote Sensing Systems QuikSCAT data. Large-scale and mesoscale characteristics of winds are compared at interannual, seasonal, and synoptic timescales. Numerical sensitivity experiments are conducted with a coupled ice-ocean model forced by different wind fields. The sensitivity experiments demonstrate differences in the net surface heat fluxes during storm events. Next, it is hypothesized that discrepancies in the wind vorticity

Magnetic susceptibility in the deep layers of the primary motor cortex in Amyotrophic Lateral Sclerosis

Directory of Open Access Journals (Sweden)

M. Costagli

2016-01-01

Full Text Available Amyotrophic Lateral Sclerosis (ALS is a progressive neurological disorder that entails degeneration of both upper and lower motor neurons. The primary motor cortex (M1 in patients with upper motor neuron (UMN impairment is pronouncedly hypointense in Magnetic Resonance (MR T2* contrast. In the present study, 3D gradient-recalled multi-echo sequences were used on a 7 Tesla MR system to acquire T2*-weighted images targeting M1 at high spatial resolution. MR raw data were used for Quantitative Susceptibility Mapping (QSM. Measures of magnetic susceptibility correlated with the expected concentration of non-heme iron in different regions of the cerebral cortex in healthy subjects. In ALS patients, significant increases in magnetic susceptibility co-localized with the T2* hypointensity observed in the middle and deep layers of M1. The magnetic susceptibility, hence iron concentration, of the deep cortical layers of patients' M1 subregions corresponding to Penfield's areas of the hand and foot in both hemispheres significantly correlated with the clinical scores of UMN impairment of the corresponding limbs. QSM therefore reflects the presence of iron deposits related to neuroinflammatory reaction and cortical microgliosis, and might prove useful in estimating M1 iron concentration, as a possible radiological sign of severe UMN burden in ALS patients.

Deep Visual Attention Prediction

Science.gov (United States)

Wang, Wenguan; Shen, Jianbing

2018-05-01

In this work, we aim to predict human eye fixation with view-free scenes based on an end-to-end deep learning architecture. Although Convolutional Neural Networks (CNNs) have made substantial improvement on human attention prediction, it is still needed to improve CNN based attention models by efficiently leveraging multi-scale features. Our visual attention network is proposed to capture hierarchical saliency information from deep, coarse layers with global saliency information to shallow, fine layers with local saliency response. Our model is based on a skip-layer network structure, which predicts human attention from multiple convolutional layers with various reception fields. Final saliency prediction is achieved via the cooperation of those global and local predictions. Our model is learned in a deep supervision manner, where supervision is directly fed into multi-level layers, instead of previous approaches of providing supervision only at the output layer and propagating this supervision back to earlier layers. Our model thus incorporates multi-level saliency predictions within a single network, which significantly decreases the redundancy of previous approaches of learning multiple network streams with different input scales. Extensive experimental analysis on various challenging benchmark datasets demonstrate our method yields state-of-the-art performance with competitive inference time.

Deep Super Learner: A Deep Ensemble for Classification Problems

OpenAIRE

Young, Steven; Abdou, Tamer; Bener, Ayse

2018-01-01

Deep learning has become very popular for tasks such as predictive modeling and pattern recognition in handling big data. Deep learning is a powerful machine learning method that extracts lower level features and feeds them forward for the next layer to identify higher level features that improve performance. However, deep neural networks have drawbacks, which include many hyper-parameters and infinite architectures, opaqueness into results, and relatively slower convergence on smaller datase...

Modeling the dispersal of Levantine Intermediate Water and its role in Mediterranean deep water formation

Science.gov (United States)

Wu, Peili; Haines, Keith

1996-03-01

This paper demonstrates the importance of Levantine Intermediate Water (LIW) in the deep water formation process in the Mediterranean using the modular ocean general circulation model at 0.25° resolution, 19 vertical levels, over the entire Mediterranean with an open Gibraltar strait. LIW formation is strongly prescribed in the Rhodes Gyre region by Haney [1971] relaxation, while in other regions, surface salinity relaxation is much reduced by applying the `mixed' thermohaline surface boundary conditions. Isopycnal diagnostics are used to trace water mass movements, and volume fluxes are monitored at straits. Low viscosity and diffusion are used to permit baroclinic eddies to play a role in water mass dispersal. The overall water budget is measured by an average flux at Gibraltar of 0.8 Sv, of which 0.7 Sv is exchanged with the eastern basin at Sicily. LIW (density around 28.95) spreads rapidly after formation throughout the entire Levantine due to baroclinic eddies. Toward the west, LIW accumulates in the northern and central Ionian, with some entering the Adriatic through Otranto and some mixing southward in eddies and exiting to the western Mediterranean through Sicily. LIW is converted to deep water in the south Adriatic at an average rate of 0.4 Sv. Water exchange through the Otranto strait appears to be buoyancy driven, with a strong bias to the end of winter (March-April), while at Sicily the exchange has a strong symmetric seasonal cycle, with maximum transport of 1.1 Sv in December indicating the effects of wind driving. LIW pathways in the west are complex and variable. In the Tyrrhenian, intermediate water becomes uniform on isopycnal surfaces due to eddy stirring. West of Sardinia, two LIW boundary currents are formed in the Balearic basin; one flows northward up the west coast of Sardinia and Corsica, and one westward along the northern African coast. The northward current is consistent with observations, while the westward current is intermittent for

Greedy Deep Dictionary Learning

OpenAIRE

Tariyal, Snigdha; Majumdar, Angshul; Singh, Richa; Vatsa, Mayank

2016-01-01

In this work we propose a new deep learning tool called deep dictionary learning. Multi-level dictionaries are learnt in a greedy fashion, one layer at a time. This requires solving a simple (shallow) dictionary learning problem, the solution to this is well known. We apply the proposed technique on some benchmark deep learning datasets. We compare our results with other deep learning tools like stacked autoencoder and deep belief network; and state of the art supervised dictionary learning t...

Chlorophyll modulation of mixed layer thermodynamics in a mixed-layer isopycnal General Circulation Model - An example from Arabian Sea and equatorial Pacific

Digital Repository Service at National Institute of Oceanography (India)

Nakamoto, S.; PrasannaKumar, S.; Oberhuber, J.M.; Saito, H.; Muneyama, K.; Frouin, R.

is influenced not only by local vertical mixing but also by horizontal con- vergence of mass and heat, a mixed layer model must consider both full dynamics due to the use of primitive equations and a parameterization for the vertical mass transfer and related... is dynamically determined without such a con- straint. Instantaneous atmospheric elds are inter- polated from the monthly means. Monthly mean climatology of chlorophyll pigment concentrations were obtained from the Coastal Zone Color Scan- ner (CZCS) from...

Learning Transferable Features with Deep Adaptation Networks

OpenAIRE

Long, Mingsheng; Cao, Yue; Wang, Jianmin; Jordan, Michael I.

2015-01-01

Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation...

Seasonal and inter-annual temperature variability in the bottom waters over the Black Sea shelf

Science.gov (United States)

Shapiro, G. I.; Wobus, F.; Aleynik, D. L.

2011-02-01

Long-term changes in the state of the Bottom Shelf Water (BSW) on the Western shelf of the Black Sea are assessed using analysis of intra- and inter-annual variations of temperature as well as their relations to physical parameters of both shelf and deep-sea waters. First, large data sets of in-situ observations over the 20th century are compiled into high-resolution monthly climatology at different depth levels. Then, the temperature anomalies from the climatic mean are calculated and aggregated into spatial compartments and seasonal bins to reveal temporal evolution of the BSW. For the purpose of this study the BSW is defined as such shelf water body between the seabed and the upper mixed layer (bounded by the σθ = 14.2 isopycnal) which has limited ability to mix vertically with oxygen-rich surface waters during the warm season (May-November) due to the formation of a seasonal pycnocline. The effects of atmospheric processes at the surface on the BSW are hence suppressed as well as the action of the "biological pump". The vertical extent of the near- bottom waters is determined based on energy considerations and the structure of the seasonal pycnocline, whilst the horizontal extent is controlled by the shelf break, where strong along-slope currents hinder exchanges with the deep sea. The BSW is shown to occupy nearly half of the area of the shelf during the summer stratification period. The potential of the BSW to ventilate horizontally during the warm season with the deep-sea waters is assessed using isopycnic analysis of temperature variations. A long-term time series of temperature anomalies in the BSW is constructed from observations during the May-November period for the 2nd half of the 20th century. The results reveal a warm phase in the 1960s/70s, followed by cooling of the BSW during 1980-2001. The transition between the warm and cold periods coincides with a regime shift in the Black Sea ecosystem. While it was confirmed that the memory of winter

Picocyanobacteria Dominance in Deep Biomass Layers: Relation to Diatom Presence and Episodic Events.

Science.gov (United States)

Aguilar, C.; Cuhel, R. L.

2016-02-01

In Offshore Marine and Large Lake Waters, most of the biomass and the productivity of phytoplankton occur below surface observation capabilities. Sub-mixed layer phytoplankton populations develop, increase, persist, and decay in relation to physical structure such as pycnocline density gradients interacting with progressively changing light fields. Basin-scale meteorological events and persistence of major invasive species have also left marks on biogeochemical cycling and ecosystem function in Lake Michigan. Among the former are precipitation and turbulence alterations brought on by unusual winter ice cover and a century-scale flood during 2008. Dampened seasonal silicate cycling indicated a basin-wide reduction of diatom production following mussel establishment. Communities in Lake Michigan shifted from diatom and big cell-dominated to small cell picocyanobacteria-dominated phytoplankton. Picocyanobacteria were beneficiaries of profound oligotrophication of the ecosystem starting in 2003. Photosynthetic parameters of pre-2003 Deep Biomass populations dominated by diatoms were systematically different from the cyanobacterial epoch following quagga mussel establishment and increase in depth of 1% incident light to 50-60m. Deep cyanobacterial production has now often been on the same scale as overlying waters. Photophysiology changes in a smooth depth gradient in this clear water as opposed to previous abrupt transition to shade adaptation. Among these many physicochemical permutations, community structure has consistently been a tradeoff between diatoms and picocyanobacteria. Opposing fluctuations of biomass favor one or the other on seasonal time frames of sequential years, with a complete system reset between each (winter mixing). For the Great Flood example, diatom surface blooms increased light extinction and drove the deep biomass maximum up - as populations settled into the pycnocline they had already outcompeted the picocyanobacteria. The opposite was true

Deep reactive ion etching of fused silica using a single-coated soft mask layer for bio-analytical applications

International Nuclear Information System (INIS)

Ray, Tathagata; Zhu, Haixin; Meldrum, Deirdre R

2010-01-01

In this note, we present our results from process development and characterization of reactive ion etching (RIE) of fused silica using a single-coated soft masking layer (KMPR® 1025, Microchem Corporation, Newton, MA). The effects of a number of fluorine-radical-based gaseous chemistries, the gas flow rate, RF power and chamber pressure on the etch rate and etching selectivity of fused silica were studied using factorial experimental designs. RF power and pressure were found to be the most important factors in determining the etch rate. The highest fused silica etch rate obtained was about 933 Å min −1 by using SF 6 -based gas chemistry, and the highest etching selectivity between the fused silica and KMPR® 1025 was up to 1.2 using a combination of CF 4 , CHF 3 and Ar. Up to 30 µm deep microstructures have been successfully fabricated using the developed processes. The average area roughness (R a ) of the etched surface was measured and results showed it is comparable to the roughness obtained using a wet etching technique. Additionally, near-vertical sidewalls (with a taper angle up to 85°) have been obtained for the etched microstructures. The processes developed here can be applied to any application requiring fabrication of deep microstructures in fused silica with near-vertical sidewalls. To our knowledge, this is the first note on deep RIE of fused silica using a single-coated KMPR® 1025 masking layer and a non-ICP-based reactive ion etcher. (technical note)

Deep Belief Networks for dimensionality reduction

NARCIS (Netherlands)

Noulas, A.K.; Kröse, B.J.A.

2008-01-01

Deep Belief Networks are probabilistic generative models which are composed by multiple layers of latent stochastic variables. The top two layers have symmetric undirected connections, while the lower layers receive directed top-down connections from the layer above. The current state-of-the-art

Deep learning

CERN Document Server

Goodfellow, Ian; Courville, Aaron

2016-01-01

Deep learning is a form of machine learning that enables computers to learn from experience and understand the world in terms of a hierarchy of concepts. Because the computer gathers knowledge from experience, there is no need for a human computer operator to formally specify all the knowledge that the computer needs. The hierarchy of concepts allows the computer to learn complicated concepts by building them out of simpler ones; a graph of these hierarchies would be many layers deep. This book introduces a broad range of topics in deep learning. The text offers mathematical and conceptual background, covering relevant concepts in linear algebra, probability theory and information theory, numerical computation, and machine learning. It describes deep learning techniques used by practitioners in industry, including deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, and practical methodology; and it surveys such applications as natural language proces...

Gas Classification Using Deep Convolutional Neural Networks

Science.gov (United States)

Peng, Pai; Zhao, Xiaojin; Pan, Xiaofang; Ye, Wenbin

2018-01-01

In this work, we propose a novel Deep Convolutional Neural Network (DCNN) tailored for gas classification. Inspired by the great success of DCNN in the field of computer vision, we designed a DCNN with up to 38 layers. In general, the proposed gas neural network, named GasNet, consists of: six convolutional blocks, each block consist of six layers; a pooling layer; and a fully-connected layer. Together, these various layers make up a powerful deep model for gas classification. Experimental results show that the proposed DCNN method is an effective technique for classifying electronic nose data. We also demonstrate that the DCNN method can provide higher classification accuracy than comparable Support Vector Machine (SVM) methods and Multiple Layer Perceptron (MLP). PMID:29316723

Gas Classification Using Deep Convolutional Neural Networks.

Science.gov (United States)

Peng, Pai; Zhao, Xiaojin; Pan, Xiaofang; Ye, Wenbin

2018-01-08

In this work, we propose a novel Deep Convolutional Neural Network (DCNN) tailored for gas classification. Inspired by the great success of DCNN in the field of computer vision, we designed a DCNN with up to 38 layers. In general, the proposed gas neural network, named GasNet, consists of: six convolutional blocks, each block consist of six layers; a pooling layer; and a fully-connected layer. Together, these various layers make up a powerful deep model for gas classification. Experimental results show that the proposed DCNN method is an effective technique for classifying electronic nose data. We also demonstrate that the DCNN method can provide higher classification accuracy than comparable Support Vector Machine (SVM) methods and Multiple Layer Perceptron (MLP).

Low absorption loss p-AlGaN superlattice cladding layer for current-injection deep ultraviolet laser diodes

Energy Technology Data Exchange (ETDEWEB)

Martens, M.; Kuhn, C.; Ziffer, E.; Simoneit, T.; Rass, J.; Wernicke, T. [Institute of Solid State Physics, Technische Universität Berlin, Hardenbergstr. 36, EW 6-1, 10623 Berlin (Germany); Kueller, V.; Knauer, A.; Einfeldt, S.; Weyers, M. [Ferdinand-Braun-Institut, Leibniz-Institut für Höchstfrequenztechnik, Gustav-Kirchhoff-Str. 4, 12489 Berlin (Germany); Kneissl, M. [Institute of Solid State Physics, Technische Universität Berlin, Hardenbergstr. 36, EW 6-1, 10623 Berlin (Germany); Ferdinand-Braun-Institut, Leibniz-Institut für Höchstfrequenztechnik, Gustav-Kirchhoff-Str. 4, 12489 Berlin (Germany)

2016-04-11

Current injection into AlGaN-based laser diode structures with high aluminum mole fractions for deep ultraviolet emission is investigated. The electrical characteristics of laser diode structures with different p-AlGaN short period superlattice (SPSL) cladding layers with various aluminum mole fractions are compared. The heterostructures contain all elements that are needed for a current-injection laser diode including cladding and waveguide layers as well as an AlGaN quantum well active region emitting near 270 nm. We found that with increasing aluminum content in the p-AlGaN cladding, the diode turn-on voltage increases, while the series resistance slightly decreases. By introducing an SPSL instead of bulk layers, the operating voltage is significantly reduced. A gain guided broad area laser diode structure with transparent p-Al{sub 0.70}Ga{sub 0.30}N waveguide layers and a transparent p-cladding with an average aluminum content of 81% was designed for strong confinement of the transverse optical mode and low optical losses. Using an optimized SPSL, this diode could sustain current densities of more than 4.5 kA/cm{sup 2}.

Low absorption loss p-AlGaN superlattice cladding layer for current-injection deep ultraviolet laser diodes

International Nuclear Information System (INIS)

Martens, M.; Kuhn, C.; Ziffer, E.; Simoneit, T.; Rass, J.; Wernicke, T.; Kueller, V.; Knauer, A.; Einfeldt, S.; Weyers, M.; Kneissl, M.

2016-01-01

Current injection into AlGaN-based laser diode structures with high aluminum mole fractions for deep ultraviolet emission is investigated. The electrical characteristics of laser diode structures with different p-AlGaN short period superlattice (SPSL) cladding layers with various aluminum mole fractions are compared. The heterostructures contain all elements that are needed for a current-injection laser diode including cladding and waveguide layers as well as an AlGaN quantum well active region emitting near 270 nm. We found that with increasing aluminum content in the p-AlGaN cladding, the diode turn-on voltage increases, while the series resistance slightly decreases. By introducing an SPSL instead of bulk layers, the operating voltage is significantly reduced. A gain guided broad area laser diode structure with transparent p-Al_0_._7_0Ga_0_._3_0N waveguide layers and a transparent p-cladding with an average aluminum content of 81% was designed for strong confinement of the transverse optical mode and low optical losses. Using an optimized SPSL, this diode could sustain current densities of more than 4.5 kA/cm"2.

Towards deep learning with segregated dendrites.

Science.gov (United States)

Guerguiev, Jordan; Lillicrap, Timothy P; Richards, Blake A

2017-12-05

Deep learning has led to significant advances in artificial intelligence, in part, by adopting strategies motivated by neurophysiology. However, it is unclear whether deep learning could occur in the real brain. Here, we show that a deep learning algorithm that utilizes multi-compartment neurons might help us to understand how the neocortex optimizes cost functions. Like neocortical pyramidal neurons, neurons in our model receive sensory information and higher-order feedback in electrotonically segregated compartments. Thanks to this segregation, neurons in different layers of the network can coordinate synaptic weight updates. As a result, the network learns to categorize images better than a single layer network. Furthermore, we show that our algorithm takes advantage of multilayer architectures to identify useful higher-order representations-the hallmark of deep learning. This work demonstrates that deep learning can be achieved using segregated dendritic compartments, which may help to explain the morphology of neocortical pyramidal neurons.

Short-term Memory of Deep RNN

OpenAIRE

Gallicchio, Claudio

2018-01-01

The extension of deep learning towards temporal data processing is gaining an increasing research interest. In this paper we investigate the properties of state dynamics developed in successive levels of deep recurrent neural networks (RNNs) in terms of short-term memory abilities. Our results reveal interesting insights that shed light on the nature of layering as a factor of RNN design. Noticeably, higher layers in a hierarchically organized RNN architecture results to be inherently biased ...

Light penetration structures the deep acoustic scattering layers in the global ocean.

KAUST Repository

Aksnes, Dag L.; Rø stad, Anders; Kaartvedt, Stein; Martinez, Udane; Duarte, Carlos M.; Irigoien, Xabier

2017-01-01

The deep scattering layer (DSL) is a ubiquitous acoustic signature found across all oceans and arguably the dominant feature structuring the pelagic open ocean ecosystem. It is formed by mesopelagic fishes and pelagic invertebrates. The DSL animals are an important food source for marine megafauna and contribute to the biological carbon pump through the active flux of organic carbon transported in their daily vertical migrations. They occupy depths from 200 to 1000 m at daytime and migrate to a varying degree into surface waters at nighttime. Their daytime depth, which determines the migration amplitude, varies across the global ocean in concert with water mass properties, in particular the oxygen regime, but the causal underpinning of these correlations has been unclear. We present evidence that the broad variability in the oceanic DSL daytime depth observed during the Malaspina 2010 Circumnavigation Expedition is governed by variation in light penetration. We find that the DSL depth distribution conforms to a common optical depth layer across the global ocean and that a correlation between dissolved oxygen and light penetration provides a parsimonious explanation for the association of shallow DSL distributions with hypoxic waters. In enhancing understanding of this phenomenon, our results should improve the ability to predict and model the dynamics of one of the largest animal biomass components on earth, with key roles in the oceanic biological carbon pump and food web.

Light penetration structures the deep acoustic scattering layers in the global ocean.

KAUST Repository

Aksnes, Dag L.

2017-05-01

The deep scattering layer (DSL) is a ubiquitous acoustic signature found across all oceans and arguably the dominant feature structuring the pelagic open ocean ecosystem. It is formed by mesopelagic fishes and pelagic invertebrates. The DSL animals are an important food source for marine megafauna and contribute to the biological carbon pump through the active flux of organic carbon transported in their daily vertical migrations. They occupy depths from 200 to 1000 m at daytime and migrate to a varying degree into surface waters at nighttime. Their daytime depth, which determines the migration amplitude, varies across the global ocean in concert with water mass properties, in particular the oxygen regime, but the causal underpinning of these correlations has been unclear. We present evidence that the broad variability in the oceanic DSL daytime depth observed during the Malaspina 2010 Circumnavigation Expedition is governed by variation in light penetration. We find that the DSL depth distribution conforms to a common optical depth layer across the global ocean and that a correlation between dissolved oxygen and light penetration provides a parsimonious explanation for the association of shallow DSL distributions with hypoxic waters. In enhancing understanding of this phenomenon, our results should improve the ability to predict and model the dynamics of one of the largest animal biomass components on earth, with key roles in the oceanic biological carbon pump and food web.

Construction of Neural Networks for Realization of Localized Deep Learning

Directory of Open Access Journals (Sweden)

Charles K. Chui

2018-05-01

Full Text Available The subject of deep learning has recently attracted users of machine learning from various disciplines, including: medical diagnosis and bioinformatics, financial market analysis and online advertisement, speech and handwriting recognition, computer vision and natural language processing, time series forecasting, and search engines. However, theoretical development of deep learning is still at its infancy. The objective of this paper is to introduce a deep neural network (also called deep-net approach to localized manifold learning, with each hidden layer endowed with a specific learning task. For the purpose of illustrations, we only focus on deep-nets with three hidden layers, with the first layer for dimensionality reduction, the second layer for bias reduction, and the third layer for variance reduction. A feedback component is also designed to deal with outliers. The main theoretical result in this paper is the order O(m-2s/(2s+d of approximation of the regression function with regularity s, in terms of the number m of sample points, where the (unknown manifold dimension d replaces the dimension D of the sampling (Euclidean space for shallow nets.

Deep and persistent melt layer in the Archaean mantle

Science.gov (United States)

Andrault, Denis; Pesce, Giacomo; Manthilake, Geeth; Monteux, Julien; Bolfan-Casanova, Nathalie; Chantel, Julien; Novella, Davide; Guignot, Nicolas; King, Andrew; Itié, Jean-Paul; Hennet, Louis

2018-02-01

The transition from the Archaean to the Proterozoic eon ended a period of great instability at the Earth's surface. The origin of this transition could be a change in the dynamic regime of the Earth's interior. Here we use laboratory experiments to investigate the solidus of samples representative of the Archaean upper mantle. Our two complementary in situ measurements of the melting curve reveal a solidus that is 200-250 K lower than previously reported at depths higher than about 100 km. Such a lower solidus temperature makes partial melting today easier than previously thought, particularly in the presence of volatiles (H2O and CO2). A lower solidus could also account for the early high production of melts such as komatiites. For an Archaean mantle that was 200-300 K hotter than today, significant melting is expected at depths from 100-150 km to more than 400 km. Thus, a persistent layer of melt may have existed in the Archaean upper mantle. This shell of molten material may have progressively disappeared because of secular cooling of the mantle. Crystallization would have increased the upper mantle viscosity and could have enhanced mechanical coupling between the lithosphere and the asthenosphere. Such a change might explain the transition from surface dynamics dominated by a stagnant lid on the early Earth to modern-like plate tectonics with deep slab subduction.

The influence of Congo River discharges in the surface and deep layers of the Gulf of Guinea

OpenAIRE

Vangriesheim, A.; Pierre, C.; Aminot, A.; Metzl, N.; Baurand, François; Caprais, J. C.

2009-01-01

The main feature of the Congo-Angola margin in the Gulf of Guinea is the Congo (ex-Zaire) deep-sea fan composed of a submarine canyon directly connected to the Congo River, a channel and a [sediment] lobe area. During the multi-disciplinary programme called BIOZAIRE conducted by Ifremer from 2000 to 2005, two CTD-O2 sections with discrete water column samples were performed (BIOZAIRE3 cruise: 2003-2004) to study the influence of the Congo River discharges, both in the surface layer and in the...

Joint Training of Deep Boltzmann Machines

OpenAIRE

Goodfellow, Ian; Courville, Aaron; Bengio, Yoshua

2012-01-01

We introduce a new method for training deep Boltzmann machines jointly. Prior methods require an initial learning pass that trains the deep Boltzmann machine greedily, one layer at a time, or do not perform well on classifi- cation tasks.

Deep Unfolding for Topic Models.

Science.gov (United States)

Chien, Jen-Tzung; Lee, Chao-Hsi

2018-02-01

Deep unfolding provides an approach to integrate the probabilistic generative models and the deterministic neural networks. Such an approach is benefited by deep representation, easy interpretation, flexible learning and stochastic modeling. This study develops the unsupervised and supervised learning of deep unfolded topic models for document representation and classification. Conventionally, the unsupervised and supervised topic models are inferred via the variational inference algorithm where the model parameters are estimated by maximizing the lower bound of logarithm of marginal likelihood using input documents without and with class labels, respectively. The representation capability or classification accuracy is constrained by the variational lower bound and the tied model parameters across inference procedure. This paper aims to relax these constraints by directly maximizing the end performance criterion and continuously untying the parameters in learning process via deep unfolding inference (DUI). The inference procedure is treated as the layer-wise learning in a deep neural network. The end performance is iteratively improved by using the estimated topic parameters according to the exponentiated updates. Deep learning of topic models is therefore implemented through a back-propagation procedure. Experimental results show the merits of DUI with increasing number of layers compared with variational inference in unsupervised as well as supervised topic models.

Deep Learning for ECG Classification

Science.gov (United States)

Pyakillya, B.; Kazachenko, N.; Mikhailovsky, N.

2017-10-01

The importance of ECG classification is very high now due to many current medical applications where this problem can be stated. Currently, there are many machine learning (ML) solutions which can be used for analyzing and classifying ECG data. However, the main disadvantages of these ML results is use of heuristic hand-crafted or engineered features with shallow feature learning architectures. The problem relies in the possibility not to find most appropriate features which will give high classification accuracy in this ECG problem. One of the proposing solution is to use deep learning architectures where first layers of convolutional neurons behave as feature extractors and in the end some fully-connected (FCN) layers are used for making final decision about ECG classes. In this work the deep learning architecture with 1D convolutional layers and FCN layers for ECG classification is presented and some classification results are showed.

Accelerating Deep Learning with Shrinkage and Recall

OpenAIRE

Zheng, Shuai; Vishnu, Abhinav; Ding, Chris

2016-01-01

Deep Learning is a very powerful machine learning model. Deep Learning trains a large number of parameters for multiple layers and is very slow when data is in large scale and the architecture size is large. Inspired from the shrinking technique used in accelerating computation of Support Vector Machines (SVM) algorithm and screening technique used in LASSO, we propose a shrinking Deep Learning with recall (sDLr) approach to speed up deep learning computation. We experiment shrinking Deep Lea...

Deep SOMs for automated feature extraction and classification from big data streaming

Science.gov (United States)

Sakkari, Mohamed; Ejbali, Ridha; Zaied, Mourad

2017-03-01

In this paper, we proposed a deep self-organizing map model (Deep-SOMs) for automated features extracting and learning from big data streaming which we benefit from the framework Spark for real time streams and highly parallel data processing. The SOMs deep architecture is based on the notion of abstraction (patterns automatically extract from the raw data, from the less to more abstract). The proposed model consists of three hidden self-organizing layers, an input and an output layer. Each layer is made up of a multitude of SOMs, each map only focusing at local headmistress sub-region from the input image. Then, each layer trains the local information to generate more overall information in the higher layer. The proposed Deep-SOMs model is unique in terms of the layers architecture, the SOMs sampling method and learning. During the learning stage we use a set of unsupervised SOMs for feature extraction. We validate the effectiveness of our approach on large data sets such as Leukemia dataset and SRBCT. Results of comparison have shown that the Deep-SOMs model performs better than many existing algorithms for images classification.

A Pacific hydrographic section at 88°W: Water-property distribution

Science.gov (United States)

Tsuchiya, Mizuki; Talley, Lynne D.

1998-06-01

Full-depth conductivity-temperature-depth (CTD)/hydrographic measurements with high horizontal and vertical resolution were made in February-April 1993 along a line lying at a nominal longitude of 88°W and extending from southern Chile (54°S) to Guatemala (14°N). It crossed five major deep basins (Southeast Pacific, Chile, Peru, Panama, and Guatemala basins) east of the East Pacific Rise. Vertical sections of potential temperature, salinity, potential density, oxygen, silica, phosphate, nitrate, and nitrite are presented to illustrate the structure of the entire water column. Some features of interest found in the sections are described, and an attempt is made to interpret them in terms of the isopycnal property distributions associated with the large-scale ocean circulation. These features include: various near-surface waters observed in the tropical and subtropical regions and the fronts that mark the boundaries of these waters; the possible importance of salt fingering to the downward salt transfer from the high-salinity subtropical water; a shallow thermostad (pycnostad) developed at 16°-18.5°C in the subtropical water; low-salinity surface water in the subantarctic zone west of southern Chile; large domains of extremely low oxygen in the subpycnocline layer on both sides of the equator and a secondary nitrite maximum associated with a nitrate minimum in these low-oxygen domains; high-salinity, low-oxygen, high-nutrient subpycnocline water that is carried poleward along the eastern boundary by the Peru-Chile Undercurrent; the Subantarctic Mode and Antarctic Intermediate waters; middepth isopycnal property extrema observed at the crest of the Sala y Gomez Ridge; influences of the North Pacific and the North Atlantic upon deep waters along the section; and the characteristics and sources of the bottom waters in the five deep basins along the section.

Unsupervised deep learning reveals prognostically relevant subtypes of glioblastoma.

Science.gov (United States)

Young, Jonathan D; Cai, Chunhui; Lu, Xinghua

2017-10-03

One approach to improving the personalized treatment of cancer is to understand the cellular signaling transduction pathways that cause cancer at the level of the individual patient. In this study, we used unsupervised deep learning to learn the hierarchical structure within cancer gene expression data. Deep learning is a group of machine learning algorithms that use multiple layers of hidden units to capture hierarchically related, alternative representations of the input data. We hypothesize that this hierarchical structure learned by deep learning will be related to the cellular signaling system. Robust deep learning model selection identified a network architecture that is biologically plausible. Our model selection results indicated that the 1st hidden layer of our deep learning model should contain about 1300 hidden units to most effectively capture the covariance structure of the input data. This agrees with the estimated number of human transcription factors, which is approximately 1400. This result lends support to our hypothesis that the 1st hidden layer of a deep learning model trained on gene expression data may represent signals related to transcription factor activation. Using the 3rd hidden layer representation of each tumor as learned by our unsupervised deep learning model, we performed consensus clustering on all tumor samples-leading to the discovery of clusters of glioblastoma multiforme with differential survival. One of these clusters contained all of the glioblastoma samples with G-CIMP, a known methylation phenotype driven by the IDH1 mutation and associated with favorable prognosis, suggesting that the hidden units in the 3rd hidden layer representations captured a methylation signal without explicitly using methylation data as input. We also found differentially expressed genes and well-known mutations (NF1, IDH1, EGFR) that were uniquely correlated with each of these clusters. Exploring these unique genes and mutations will allow us to

The application of deep confidence network in the problem of image recognition

Directory of Open Access Journals (Sweden)

Chumachenko О.І.

2016-12-01

Full Text Available In order to study the concept of deep learning, in particular the substitution of multilayer perceptron on the corresponding network of deep confidence, computer simulations of the learning process to test voters was carried out. Multi-layer perceptron has been replaced by a network of deep confidence, consisting of successive limited Boltzmann machines. After training of a network of deep confidence algorithm of layer-wise training it was found that the use of networks of deep confidence greatly improves the accuracy of multilayer perceptron training by method of reverse distribution errors.

Deep kernel learning method for SAR image target recognition

Science.gov (United States)

Chen, Xiuyuan; Peng, Xiyuan; Duan, Ran; Li, Junbao

2017-10-01

With the development of deep learning, research on image target recognition has made great progress in recent years. Remote sensing detection urgently requires target recognition for military, geographic, and other scientific research. This paper aims to solve the synthetic aperture radar image target recognition problem by combining deep and kernel learning. The model, which has a multilayer multiple kernel structure, is optimized layer by layer with the parameters of Support Vector Machine and a gradient descent algorithm. This new deep kernel learning method improves accuracy and achieves competitive recognition results compared with other learning methods.

Radium 228 as a tracer of basin wide processes in the Abyssal Ocean

International Nuclear Information System (INIS)

Sarmiento, J.L.; Rooth, C.G.H.; Broecker, W.S.

1982-01-01

A simple model of isopycnal mixing in a circular basin is developed in order to examine the utility of the 5.75-year half-life tracer radium 228 for studying basin wide processes in the deep ocean. The model shows that it is possible to resolve diffusivities of 7 cm 2 s - 1 in a basin of approx.3000-km diameter with profiles measured near the center and edge of the basin. A least squares fit of the model to four abyssal profiles measured during GEOSECS in the North American Basin gives an isopycnal diffusivity of 6 x 10 7 cm 2 s - 1

Deep levels in silicon–oxygen superlattices

International Nuclear Information System (INIS)

Simoen, E; Jayachandran, S; Delabie, A; Caymax, M; Heyns, M

2016-01-01

This work reports on the deep levels observed in Pt/Al 2 O 3 /p-type Si metal-oxide-semiconductor capacitors containing a silicon–oxygen superlattice (SL) by deep-level transient spectroscopy. It is shown that the presence of the SL gives rise to a broad band of hole traps occurring around the silicon mid gap, which is absent in reference samples with a silicon epitaxial layer. In addition, the density of states of the deep layers roughly scales with the number of SL periods for the as-deposited samples. Annealing in a forming gas atmosphere reduces the maximum concentration significantly, while the peak energy position shifts from close-to mid-gap towards the valence band edge. Based on the flat-band voltage shift of the Capacitance–Voltage characteristics it is inferred that positive charge is introduced by the oxygen atomic layers in the SL, indicating the donor nature of the underlying hole traps. In some cases, a minor peak associated with P b dangling bond centers at the Si/SiO 2 interface has been observed as well. (paper)

Refinement of learned skilled movement representation in motor cortex deep output layer

Science.gov (United States)

Li, Qian; Ko, Ho; Qian, Zhong-Ming; Yan, Leo Y. C.; Chan, Danny C. W.; Arbuthnott, Gordon; Ke, Ya; Yung, Wing-Ho

2017-01-01

The mechanisms underlying the emergence of learned motor skill representation in primary motor cortex (M1) are not well understood. Specifically, how motor representation in the deep output layer 5b (L5b) is shaped by motor learning remains virtually unknown. In rats undergoing motor skill training, we detect a subpopulation of task-recruited L5b neurons that not only become more movement-encoding, but their activities are also more structured and temporally aligned to motor execution with a timescale of refinement in tens-of-milliseconds. Field potentials evoked at L5b in vivo exhibit persistent long-term potentiation (LTP) that parallels motor performance. Intracortical dopamine denervation impairs motor learning, and disrupts the LTP profile as well as the emergent neurodynamical properties of task-recruited L5b neurons. Thus, dopamine-dependent recruitment of L5b neuronal ensembles via synaptic reorganization may allow the motor cortex to generate more temporally structured, movement-encoding output signal from M1 to downstream circuitry that drives increased uniformity and precision of movement during motor learning. PMID:28598433

Deep Support Vector Machines for Regression Problems

NARCIS (Netherlands)

Wiering, Marco; Schutten, Marten; Millea, Adrian; Meijster, Arnold; Schomaker, Lambertus

2013-01-01

In this paper we describe a novel extension of the support vector machine, called the deep support vector machine (DSVM). The original SVM has a single layer with kernel functions and is therefore a shallow model. The DSVM can use an arbitrary number of layers, in which lower-level layers contain

Shear Strengthening of RC Deep Beam Using Externally Bonded GFRP Fabrics

Science.gov (United States)

Kumari, A.; Patel, S. S.; Nayak, A. N.

2018-06-01

This work presents the experimental investigation of RC deep beams wrapped with externally bonded Glass Fibre Reinforced Polymer (GFRP) fabrics in order to study the Load versus deflection behavior, cracking pattern, failure modes and ultimate shear strength. A total number of five deep beams have been casted, which is designed with conventional steel reinforcement as per IS: 456 (Indian standard plain and reinforced concrete—code for practice, Bureau of Indian Standards, New Delhi, 2000). The spans to depth ratio for all RC deep beams have been kept less than 2 as per the above specification. Out of five RC deep beams, one without retrofitting serves as a reference beam and the rest four have been wrapped with GFRP fabrics in multiple layers and tested with two point loading condition. The first cracking load, ultimate load and the shear contribution of GFRP to the deep beams have been observed. A critical discussion is made with respect to the enhancement of the strength, behaviour and performance of retrofitted deep beams in comparison to the deep beam without GFRP in order to explore the potential use of GFRP for strengthening the RC deep beams. Test results have demonstrated that the deep beams retrofitted with GFRP shows a slower development of the diagonal cracks and improves shear carrying capacity of the RC deep beam. A comparative study of the experimental results with the theoretical ones predicted by various researchers available in the literatures has also been presented. It is observed that the ultimate load of the beams retrofitted with GFRP fabrics increases with increase of number of GFRP layers up to a specific number of layers, i.e. 3 layers, beyond which it decreases.

Seasonal evolution of physical processes and biological responses in the northern Red Sea

KAUST Repository

Asfahani, Khaled

2017-12-01

A sequence of autonomous underwater glider deployments were used to characterize the spatial-temporal variability of the region over an eight month period from late September to May. Strongly stratified system was found in early fall with significant gradients in both temperature (T) and salinity (S), during winter T < 23°C and minimum S of 40.3 psu was observed and resulting in weakened stratification that enables deep convective mixing and upwelling of deep water by cyclonic circulations in the region leading to significant biomass increase. Throughout the entire observational period the slope of the 28 and 28.5 kg/m3 isopycnals remained sloping downward from offshore toward the coast reflected a persistent northward geostrophic flow. The depth of the 180 μmol/kg isopleth of oxygen, indicative of the top of the nutricline, paralleled the depth of the 28 kg/m3, but remained slightly deeper than the isopycnal. The deep winter mixing did not penetrate the nutricline where the mixed layer was deeper near the coast. However, because of the cyclonic signature the 28 kg/m3 rose to the surface offshore, injecting nutrients into the surface layer and promoting increased biomass in the central Red Sea. With the presence of cyclonic eddies, there was evidence of subduction associated with the cross-eddy circulation. This subducted flow was toward the coast within the domain of the glider observations. During this period, increases in the particulate backscatter were associated with increased chlorophyll indicating that the suspended particles were primarily phytoplankton particles. Within the mean northward flow there is a cross-basin flow wherein water is upwelled near the center of the Red Sea, there is a eastward component to the northward flow, and subsequent downwelling near the coasts. Within the surface flow subductive processes lead not only to a horizontal flow, but also a downward component toward the coast. Overall transport is very 3-dimensional in the

Auxiliary Deep Generative Models

DEFF Research Database (Denmark)

Maaløe, Lars; Sønderby, Casper Kaae; Sønderby, Søren Kaae

2016-01-01

Deep generative models parameterized by neural networks have recently achieved state-of-the-art performance in unsupervised and semi-supervised learning. We extend deep generative models with auxiliary variables which improves the variational approximation. The auxiliary variables leave...... the generative model unchanged but make the variational distribution more expressive. Inspired by the structure of the auxiliary variable we also propose a model with two stochastic layers and skip connections. Our findings suggest that more expressive and properly specified deep generative models converge...... faster with better results. We show state-of-the-art performance within semi-supervised learning on MNIST (0.96%), SVHN (16.61%) and NORB (9.40%) datasets....

Comparison of frailty of primary neurons, embryonic, and aging mouse cortical layers.

Science.gov (United States)

Fugistier, Patrick; Vallet, Philippe G; Leuba, Geneviève; Piotton, Françoise; Marin, Pascale; Bouras, Constantin; Savioz, Armand

2014-02-01

Superficial layers I to III of the human cerebral cortex are more vulnerable toward Aβ peptides than deep layers V to VI in aging. Three models of layers were used to investigate this pattern of frailty. First, primary neurons from E14 and E17 embryonic murine cortices, corresponding respectively to future deep and superficial layers, were treated either with Aβ(1-42), okadaic acid, or kainic acid. Second, whole E14 and E17 embryonic cortices, and third, in vitro separated deep and superficial layers of young and old C57BL/6J mice, were treated identically. We observed that E14 and E17 neurons in culture were prone to death after the Aβ and particularly the kainic acid treatment. This was also the case for the superficial layers of the aged cortex, but not for the embryonic, the young cortex, and the deep layers of the aged cortex. Thus, the aged superficial layers appeared to be preferentially vulnerable against Aβ and kainic acid. This pattern of vulnerability corresponds to enhanced accumulation of senile plaques in the superficial cortical layers with aging and Alzheimer's disease. Copyright © 2014 Elsevier Inc. All rights reserved.

Temperature and moisture effects on greenhouse gas emissions from deep active-layer boreal soils

Science.gov (United States)

Bond-Lamberty, Ben; Smith, A. Peyton; Bailey, Vanessa

2016-12-01

Rapid climatic changes, rising air temperatures, and increased fires are expected to drive permafrost degradation and alter soil carbon (C) cycling in many high-latitude ecosystems. How these soils will respond to changes in their temperature, moisture, and overlying vegetation is uncertain but critical to understand given the large soil C stocks in these regions. We used a laboratory experiment to examine how temperature and moisture control CO2 and CH4 emissions from mineral soils sampled from the bottom of the annual active layer, i.e., directly above permafrost, in an Alaskan boreal forest. Gas emissions from 30 cores, subjected to two temperatures and either field moisture conditions or experimental drought, were tracked over a 100-day incubation; we also measured a variety of physical and chemical characteristics of the cores. Gravimetric water content was 0.31 ± 0.12 (unitless) at the beginning of the incubation; cores at field moisture were unchanged at the end, but drought cores had declined to 0.06 ± 0.04. Daily CO2 fluxes were positively correlated with incubation chamber temperature, core water content, and percent soil nitrogen. They also had a temperature sensitivity (Q10) of 1.3 and 1.9 for the field moisture and drought treatments, respectively. Daily CH4 emissions were most strongly correlated with percent nitrogen, but neither temperature nor water content was a significant first-order predictor of CH4 fluxes. The cumulative production of C from CO2 was over 6 orders of magnitude higher than that from CH4; cumulative CO2 was correlated with incubation temperature and moisture treatment, with drought cores producing 52-73 % lower C. Cumulative CH4 production was unaffected by any treatment. These results suggest that deep active-layer soils may be sensitive to changes in soil moisture under aerobic conditions, a critical factor as discontinuous permafrost thaws in interior Alaska. Deep but unfrozen high-latitude soils have been shown to be

Laminar recordings in frontal cortex suggest distinct layers for maintenance and control of working memory.

Science.gov (United States)

Bastos, André M; Loonis, Roman; Kornblith, Simon; Lundqvist, Mikael; Miller, Earl K

2018-01-30

All of the cerebral cortex has some degree of laminar organization. These different layers are composed of neurons with distinct connectivity patterns, embryonic origins, and molecular profiles. There are little data on the laminar specificity of cognitive functions in the frontal cortex, however. We recorded neuronal spiking/local field potentials (LFPs) using laminar probes in the frontal cortex (PMd, 8A, 8B, SMA/ACC, DLPFC, and VLPFC) of monkeys performing working memory (WM) tasks. LFP power in the gamma band (50-250 Hz) was strongest in superficial layers, and LFP power in the alpha/beta band (4-22 Hz) was strongest in deep layers. Memory delay activity, including spiking and stimulus-specific gamma bursting, was predominately in superficial layers. LFPs from superficial and deep layers were synchronized in the alpha/beta bands. This was primarily unidirectional, with alpha/beta bands in deep layers driving superficial layer activity. The phase of deep layer alpha/beta modulated superficial gamma bursting associated with WM encoding. Thus, alpha/beta rhythms in deep layers may regulate the superficial layer gamma bands and hence maintenance of the contents of WM. Copyright © 2018 the Author(s). Published by PNAS.

Deep water characteristics and circulation in the South China Sea

Science.gov (United States)

Wang, Aimei; Du, Yan; Peng, Shiqiu; Liu, Kexiu; Huang, Rui Xin

2018-04-01

This study investigates the deep circulation in the South China Sea (SCS) using oceanographic observations combined with results from a bottom layer reduced gravity model. The SCS water, 2000 m below the surface, is quite different from that in the adjacent Pacific Ocean, and it is characterized by its low dissolved oxygen (DO), high temperature and low salinity. The horizontal distribution of deep water properties indicates a basin-scale cyclonic circulation driven by the Luzon overflow. The results of the bottom layer reduced gravity model are consistent with the existence of the cyclonic circulation in the deep SCS. The circulation is stronger at the northern/western boundary. After overflowing the sill of the Luzon Strait, the deep water moves broadly southwestward, constrained by the 3500 m isobath. The broadening of the southward flow is induced by the downwelling velocity in the interior of the deep basin. The main deep circulation bifurcates into two branches after the Zhongsha Islands. The southward branch continues flowing along the 3500 m isobath, and the eastward branch forms the sub-basin scale cyclonic circulation around the seamounts in the central deep SCS. The returning flow along the east boundary is fairly weak. The numerical experiments of the bottom layer reduced gravity model reveal the important roles of topography, bottom friction, and the upwelling/downwelling pattern in controlling the spatial structure, particularly the strong, deep western boundary current.

A simple model of the effect of ocean ventilation on ocean heat uptake

Energy Technology Data Exchange (ETDEWEB)

Nadiga, Balasubramanya T. [Los Alamos National Lab. (LANL), Los Alamos, NM (United States); Urban, Nathan Mark [Los Alamos National Lab. (LANL), Los Alamos, NM (United States)

2018-01-27

Presentation includes slides on Earth System Models vs. Simple Climate Models; A Popular SCM: Energy Balance Model of Anomalies; On calibrating against one ESM experiment, the SCM correctly captures that ESM's surface warming response with other forcings; Multi-Model Analysis: Multiple ESMs, Single SCM; Posterior Distributions of ECS; However In Excess of 90% of TOA Energy Imbalance is Sequestered in the World Oceans; Heat Storage in the Two Layer Model; Heat Storage in the Two Layer Model; Including TOA Rad. Imbalance and Ocean Heat in Calibration Improves Repr., but Significant Errors Persist; Improved Vertical Resolution Does Not Fix Problem; A Series of Expts. Confirms That Anomaly-Diffusing Models Cannot Properly Represent Ocean Heat Uptake; Physics of the Thermocline; Outcropping Isopycnals and Horizontally-Averaged Layers; Local interactions between outcropping isopycnals leads to non-local interactions between horizontally-averaged layers; Both Surface Warming and Ocean Heat are Well Represented With Just 4 Layers; A Series of Expts. Confirms That When Non-Local Interactions are Allowed, the SCMs Can Represent Both Surface Warming and Ocean Heat Uptake; and Summary and Conclusions.

pDeep: Predicting MS/MS Spectra of Peptides with Deep Learning.

Science.gov (United States)

Zhou, Xie-Xuan; Zeng, Wen-Feng; Chi, Hao; Luo, Chunjie; Liu, Chao; Zhan, Jianfeng; He, Si-Min; Zhang, Zhifei

2017-12-05

In tandem mass spectrometry (MS/MS)-based proteomics, search engines rely on comparison between an experimental MS/MS spectrum and the theoretical spectra of the candidate peptides. Hence, accurate prediction of the theoretical spectra of peptides appears to be particularly important. Here, we present pDeep, a deep neural network-based model for the spectrum prediction of peptides. Using the bidirectional long short-term memory (BiLSTM), pDeep can predict higher-energy collisional dissociation, electron-transfer dissociation, and electron-transfer and higher-energy collision dissociation MS/MS spectra of peptides with >0.9 median Pearson correlation coefficients. Further, we showed that intermediate layer of the neural network could reveal physicochemical properties of amino acids, for example the similarities of fragmentation behaviors between amino acids. We also showed the potential of pDeep to distinguish extremely similar peptides (peptides that contain isobaric amino acids, for example, GG = N, AG = Q, or even I = L), which were very difficult to distinguish using traditional search engines.

DeepNAT: Deep convolutional neural network for segmenting neuroanatomy.

Science.gov (United States)

Wachinger, Christian; Reuter, Martin; Klein, Tassilo

2018-04-15

We introduce DeepNAT, a 3D Deep convolutional neural network for the automatic segmentation of NeuroAnaTomy in T1-weighted magnetic resonance images. DeepNAT is an end-to-end learning-based approach to brain segmentation that jointly learns an abstract feature representation and a multi-class classification. We propose a 3D patch-based approach, where we do not only predict the center voxel of the patch but also neighbors, which is formulated as multi-task learning. To address a class imbalance problem, we arrange two networks hierarchically, where the first one separates foreground from background, and the second one identifies 25 brain structures on the foreground. Since patches lack spatial context, we augment them with coordinates. To this end, we introduce a novel intrinsic parameterization of the brain volume, formed by eigenfunctions of the Laplace-Beltrami operator. As network architecture, we use three convolutional layers with pooling, batch normalization, and non-linearities, followed by fully connected layers with dropout. The final segmentation is inferred from the probabilistic output of the network with a 3D fully connected conditional random field, which ensures label agreement between close voxels. The roughly 2.7million parameters in the network are learned with stochastic gradient descent. Our results show that DeepNAT compares favorably to state-of-the-art methods. Finally, the purely learning-based method may have a high potential for the adaptation to young, old, or diseased brains by fine-tuning the pre-trained network with a small training sample on the target application, where the availability of larger datasets with manual annotations may boost the overall segmentation accuracy in the future. Copyright © 2017 Elsevier Inc. All rights reserved.

The deep chlorophyll layer in Lake Ontario: Extent, mechanisms of formation, and abiotic predictors

Science.gov (United States)

Scofield, Anne E.; Watkins, James M.; Weidel, Brian C.; Luckey, Frederick J.; Rudstam, Lars G.

2017-01-01

Epilimnetic production has declined in Lake Ontario, but increased production in metalimnetic deep chlorophyll layers (DCLs) may compensate for these losses. We investigated the spatial and temporal extent of DCLs, the mechanisms driving DCL formation, and the use of physical variables for predicting the depth and concentration of the deep chlorophyll maximum (DCM) during April–September 2013. A DCL with DCM concentrations 2 to 3 times greater than those in the epilimnion was present when the euphotic depth extended below the epilimnion, which occurred primarily from late June through mid-August. In situ growth was important for DCL formation in June and July, but settling and photoadaptation likely also contributed to the later-season DCL. Supporting evidence includes: phytoplankton biovolume was 2.4 × greater in the DCL than in the epilimnion during July, the DCL phytoplankton community of July was different from that of May and the July epilimnion (p = 0.004), and there were concurrences of DCM with maxima in fine particle concentration and dissolved oxygen saturation. Higher nutrient levels in the metalimnion may also be a necessary condition for DCL formation because July metalimnetic concentrations were 1.5 × (nitrate) and 3.5 × (silica) greater than in the epilimnion. Thermal structure variables including epilimnion depth, thermocline depth, and thermocline steepness were useful for predicting DCM depth; the inclusion of euphotic depth only marginally improved these predictions. However, euphotic depth was critical for predicting DCM concentrations. The DCL is a productive and predictable feature of the Lake Ontario ecosystem during the stratified period.

Convectively-driven cold layer and its influences on moisture in the UTLS

Science.gov (United States)

Kim, J.; Randel, W. J.; Birner, T.

2016-12-01

Characteristics of the cold anomaly in the tropical tropopause layer (TTL) that is commonly observed with deep convection are examined using CloudSat and Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC) GPS radio occultation measurements. Deep convection is sampled based on the cloud top height (>17 km) from CloudSat 2B-CLDCLASS, and then temperature profiles from COSMIC are composited around the deep convection. The composite temperature shows anomalously warm troposphere (up to 14 km) and a significantly cold layer near the tropopause (at 16-18 km) in the regions of deep convection. Generally in the tropics, the cold layer has very large horizontal scale (2,000 - 6,000 km) compared to that of mesoscale convective cluster, and it lasts one or two weeks with minimum temperature anomaly of - 2K. The cold layer shows slight but clear eastward-tilted vertical structure in the deep tropics indicating a large-scale Kelvin wave response. Further analyses on circulation patterns suggest that the anomaly can be explained as a part of Gill-type response in the TTL to deep convective heating in the troposphere. Response of moisture to the cold layer is also examined in the upper troposphere and lower stratosphere using microwave limb sounder (MLS) measurements. The water vapor anomalies show coherent structures with the temperature and circulation anomalies. A clear dry anomaly is found in the cold layer and its outflow region, implying a large-scale dehydration process due to the convectively driven cold layer in the upper TTL.

Tracing Internal Radar Layers in the Greenland Ice Sheet

DEFF Research Database (Denmark)

Panton, Christian

Internal layers in radio-echograms from the sounding of ice sheets have long been a valuable resource in glaciology, but their usefulness have been limited by availability of traced (digitized) layers. To speed up this process, we have developed an algorithm for semi-automatic tracing the internal...... layers and a fully automated algorithm for mapping the layer slope. The layer slope is inferred by the intensity response to a slanted Gaussian filter, whereafter layers can be traced using an active contour model. With these techniques we show that it possible to trace internal layers over distances...... of hundreds kilometers with minimal operator intervention, and the methods have been successfully validated between two Greenland deep ice cores with internal match points. In order to remove any operator assistance, we show how the layer slope can be used to detect disturbances in the deep radiostratigraphy...

Mixed layer depth trends in the Bay of Biscay over the period 1975-2010.

Directory of Open Access Journals (Sweden)

Xurxo Costoya

Full Text Available Wintertime trends in mixed layer depth (MLD were calculated in the Bay of Biscay over the period 1975-2010 using the Simple Ocean Data Assimilation (SODA package. The reliability of the SODA database was confirmed correlating its results with those obtained from the experimental Argo database over the period 2003-2010. An iso-thermal layer depth (TLD and an iso-pycnal layer depth (PLD were defined using the threshold difference method with ΔT = 0.5°C and Δσθ = 0.125 kg/m3. Wintertime trends of the MLD were calculated using winter extended (December-March anomalies and annual maxima. Trends calculated for the whole Bay of Biscay using both parameters (TLD and PLD showed to be dependent on the area. Thus, MLD became deeper in the southeastern corner and shallower in the rest of the area. Air temperature was shown to play a key role in regulating the different spatial behavior of the MLD. Negative air temperature trends localized in the southeastern corner coincide with MLD deepening in this area, while, positive air temperature trends are associated to MLD shoaling in the rest of the bay. Additionally, the temperature trend calculated along the first 700 m of the water column is in good agreement with the different spatial behavior revealed for the MLD trend.

Deep levels in metamorphic InAs/InGaAs quantum dot structures with different composition of the embedding layers

Science.gov (United States)

Golovynskyi, S.; Datsenko, O.; Seravalli, L.; Kozak, O.; Trevisi, G.; Frigeri, P.; Babichuk, I. S.; Golovynska, I.; Qu, Junle

2017-12-01

Deep levels in metamorphic InAs/In x Ga1-x As quantum dot (QD) structures are studied with deep level thermally stimulated conductivity (TSC), photoconductivity (PC) and photoluminescence (PL) spectroscopy and compared with data from pseudomorphic InGaAs/GaAs QDs investigated previously by the same techniques. We have found that for a low content of indium (x = 0.15) the trap density in the plane of self-assembled QDs is comparable or less than the one for InGaAs/GaAs QDs. However, the trap density increases with x, resulting in a rise of the defect photoresponse in PC and TSC spectra as well as a reduction of the QD PL intensity. The activation energies of the deep levels and some traps correspond to known defect complexes EL2, EL6, EL7, EL9, and EL10 inherent in GaAs, and three traps are attributed to the extended defects, located in InGaAs embedding layers. The rest of them have been found as concentrated mainly close to QDs, as their density in the deeper InGaAs buffers is much lower. This an important result for the development of light-emitting and light-sensitive devices based on metamorphic InAs QDs, as it is a strong indication that the defect density is not higher than in pseudomorphic InAs QDs.

Invited talk: Deep Learning Meets Physics

CERN Multimedia

CERN. Geneva

2018-01-01

Deep Learning has emerged as one of the most successful fields of machine learning and artificial intelligence with overwhelming success in industrial speech, text and vision benchmarks. Consequently it evolved into the central field of research for IT giants like Google, facebook, Microsoft, Baidu, and Amazon. Deep Learning is founded on novel neural network techniques, the recent availability of very fast computers, and massive data sets. In its core, Deep Learning discovers multiple levels of abstract representations of the input. The main obstacle to learning deep neural networks is the vanishing gradient problem. The vanishing gradient impedes credit assignment to the first layers of a deep network or to early elements of a sequence, therefore limits model selection. Major advances in Deep Learning can be related to avoiding the vanishing gradient like stacking, ReLUs, residual networks, highway networks, and LSTM. For Deep Learning, we suggested self-normalizing neural networks (SNNs) which automatica...

Classification of Exacerbation Frequency in the COPDGene Cohort Using Deep Learning with Deep Belief Networks.

Science.gov (United States)

Ying, Jun; Dutta, Joyita; Guo, Ning; Hu, Chenhui; Zhou, Dan; Sitek, Arkadiusz; Li, Quanzheng

2016-12-21

This study aims to develop an automatic classifier based on deep learning for exacerbation frequency in patients with chronic obstructive pulmonary disease (COPD). A threelayer deep belief network (DBN) with two hidden layers and one visible layer was employed to develop classification models and the models' robustness to exacerbation was analyzed. Subjects from the COPDGene cohort were labeled with exacerbation frequency, defined as the number of exacerbation events per year. 10,300 subjects with 361 features each were included in the analysis. After feature selection and parameter optimization, the proposed classification method achieved an accuracy of 91.99%, using a 10-fold cross validation experiment. The analysis of DBN weights showed that there was a good visual spatial relationship between the underlying critical features of different layers. Our findings show that the most sensitive features obtained from the DBN weights are consistent with the consensus showed by clinical rules and standards for COPD diagnostics. We thus demonstrate that DBN is a competitive tool for exacerbation risk assessment for patients suffering from COPD.

Impact of deep convection in the tropical tropopause layer in West Africa: in-situ observations and mesoscale modelling

Directory of Open Access Journals (Sweden)

F. Fierli

2011-01-01

Full Text Available We present the analysis of the impact of convection on the composition of the tropical tropopause layer region (TTL in West-Africa during the AMMA-SCOUT campaign. Geophysica M55 aircraft observations of water vapor, ozone, aerosol and CO₂ during August 2006 show perturbed values at altitudes ranging from 14 km to 17 km (above the main convective outflow and satellite data indicates that air detrainment is likely to have originated from convective cloud east of the flights. Simulations of the BOLAM mesoscale model, nudged with infrared radiance temperatures, are used to estimate the convective impact in the upper troposphere and to assess the fraction of air processed by convection. The analysis shows that BOLAM correctly reproduces the location and the vertical structure of convective outflow. Model-aided analysis indicates that convection can influence the composition of the upper troposphere above the level of main outflow for an event of deep convection close to the observation site. Model analysis also shows that deep convection occurring in the entire Sahelian transect (up to 2000 km E of the measurement area has a non negligible role in determining TTL composition.

Sacrificial wafer bonding for planarization after very deep etching

NARCIS (Netherlands)

Spiering, V.L.; Spiering, Vincent L.; Berenschot, Johan W.; Elwenspoek, Michael Curt; Fluitman, J.H.J.

A new technique is presented that provides planarization after a very deep etching step in silicon. This offers the possibility for as well resist spinning and layer patterning as realization of bridges or cantilevers across deep holes or grooves. The sacrificial wafer bonding technique contains a

IMPROVEMENT OF RECOGNITION QUALITY IN DEEP LEARNING NETWORKS BY SIMULATED ANNEALING METHOD

Directory of Open Access Journals (Sweden)

A. S. Potapov

2014-09-01

Full Text Available The subject of this research is deep learning methods, in which automatic construction of feature transforms is taken place in tasks of pattern recognition. Multilayer autoencoders have been taken as the considered type of deep learning networks. Autoencoders perform nonlinear feature transform with logistic regression as an upper classification layer. In order to verify the hypothesis of possibility to improve recognition rate by global optimization of parameters for deep learning networks, which are traditionally trained layer-by-layer by gradient descent, a new method has been designed and implemented. The method applies simulated annealing for tuning connection weights of autoencoders while regression layer is simultaneously trained by stochastic gradient descent. Experiments held by means of standard MNIST handwritten digit database have shown the decrease of recognition error rate from 1.1 to 1.5 times in case of the modified method comparing to the traditional method, which is based on local optimization. Thus, overfitting effect doesn’t appear and the possibility to improve learning rate is confirmed in deep learning networks by global optimization methods (in terms of increasing recognition probability. Research results can be applied for improving the probability of pattern recognition in the fields, which require automatic construction of nonlinear feature transforms, in particular, in the image recognition. Keywords: pattern recognition, deep learning, autoencoder, logistic regression, simulated annealing.

Diffusive boundary layers over varying topography

KAUST Repository

Dell, R. W.

2015-03-25

Diffusive bottom boundary layers can produce upslope flows in a stratified fluid. Accumulating observations suggest that these boundary layers may drive upwelling and mixing in mid-ocean ridge flank canyons. However, most studies of diffusive bottom boundary layers to date have concentrated on constant bottom slopes. We present a study of how diffusive boundary layers interact with various idealized topography, such as changes in bottom slope, slopes with corrugations and isolated sills. We use linear theory and numerical simulations in the regional ocean modeling system (ROMS) model to show changes in bottom slope can cause convergences and divergences within the boundary layer, in turn causing fluid exchanges that reach far into the overlying fluid and alter stratification far from the bottom. We also identify several different regimes of boundary-layer behaviour for topography with oceanographically relevant size and shape, including reversing flows and overflows, and we develop a simple theory that predicts the regime boundaries, including what topographies will generate overflows. As observations also suggest there may be overflows in deep canyons where the flow passes over isolated bumps and sills, this parameter range may be particularly significant for understanding the role of boundary layers in the deep ocean.

Exploring Ocean Animal Trajectory Pattern via Deep Learning

KAUST Repository

Wang, Su

2016-01-01

We trained a combined deep convolutional neural network to predict seals’ age (3 categories) and gender (2 categories). The entire dataset contains 110 seals with around 489 thousand location records. Most records are continuous and measured in a certain step. We created five convolutional layers for feature representation and established two fully connected structure as age’s and gender’s classifier, respectively. Each classifier consists of three fully connected layers. Treating seals’ latitude and longitude as input, entire deep learning network, which includes 780,000 neurons and 2,097,000 parameters, can reach to 70.72% accuracy rate for predicting seals’ age and simultaneously achieve 79.95% for gender estimation.

Exploring Ocean Animal Trajectory Pattern via Deep Learning

KAUST Repository

Wang, Su

2016-05-23

We trained a combined deep convolutional neural network to predict seals’ age (3 categories) and gender (2 categories). The entire dataset contains 110 seals with around 489 thousand location records. Most records are continuous and measured in a certain step. We created five convolutional layers for feature representation and established two fully connected structure as age’s and gender’s classifier, respectively. Each classifier consists of three fully connected layers. Treating seals’ latitude and longitude as input, entire deep learning network, which includes 780,000 neurons and 2,097,000 parameters, can reach to 70.72% accuracy rate for predicting seals’ age and simultaneously achieve 79.95% for gender estimation.

Active semi-supervised learning method with hybrid deep belief networks.

Science.gov (United States)

Zhou, Shusen; Chen, Qingcai; Wang, Xiaolong

2014-01-01

In this paper, we develop a novel semi-supervised learning algorithm called active hybrid deep belief networks (AHD), to address the semi-supervised sentiment classification problem with deep learning. First, we construct the previous several hidden layers using restricted Boltzmann machines (RBM), which can reduce the dimension and abstract the information of the reviews quickly. Second, we construct the following hidden layers using convolutional restricted Boltzmann machines (CRBM), which can abstract the information of reviews effectively. Third, the constructed deep architecture is fine-tuned by gradient-descent based supervised learning with an exponential loss function. Finally, active learning method is combined based on the proposed deep architecture. We did several experiments on five sentiment classification datasets, and show that AHD is competitive with previous semi-supervised learning algorithm. Experiments are also conducted to verify the effectiveness of our proposed method with different number of labeled reviews and unlabeled reviews respectively.

Structure and Variability of Internal Tides in Luzon Strait

Science.gov (United States)

2016-09-14

velocity, and (e) S9 meridional velocity. Black lines are isopycnal contours at approximately 500 -m intervals. Contour values (kgm23) are (1023.2, 1026.8...W., and C. Wunsch, 1998: Abyssal recipes II: Energetics of tidal and wind mixing. Deep-Sea Res. I, 45, 1977–2010, doi:10.1016/S0967-0637(98)00070-3

Deep Learning from Crowds

DEFF Research Database (Denmark)

Rodrigues, Filipe; Pereira, Francisco Camara

Over the last few years, deep learning has revolutionized the field of machine learning by dramatically improving the stateof-the-art in various domains. However, as the size of supervised artificial neural networks grows, typically so does the need for larger labeled datasets. Recently...... networks from crowds. We begin by describing an EM algorithm for jointly learning the parameters of the network and the reliabilities of the annotators. Then, a novel general-purpose crowd layer is proposed, which allows us to train deep neural networks end-to-end, directly from the noisy labels......, crowdsourcing has established itself as an efficient and cost-effective solution for labeling large sets of data in a scalable manner, but it often requires aggregating labels from multiple noisy contributors with different levels of expertise. In this paper, we address the problem of learning deep neural...

Enhanced sensitivity of DNA- and rRNA-based stable isotope probing by fractionation and quantitative analysis of isopycnic centrifugation gradients.

Science.gov (United States)

Lueders, Tillmann; Manefield, Mike; Friedrich, Michael W

2004-01-01

Stable isotope probing (SIP) of nucleic acids allows the detection and identification of active members of natural microbial populations that are involved in the assimilation of an isotopically labelled compound into nucleic acids. SIP is based on the separation of isotopically labelled DNA or rRNA by isopycnic density gradient centrifugation. We have developed a highly sensitive protocol for the detection of 'light' and 'heavy' nucleic acids in fractions of centrifugation gradients. It involves the fluorometric quantification of total DNA or rRNA, and the quantification of either 16S rRNA genes or 16S rRNA in gradient fractions by real-time PCR with domain-specific primers. Using this approach, we found that fully 13C-labelled DNA or rRNA of Methylobacterium extorquens was quantitatively resolved from unlabelled DNA or rRNA of Methanosarcina barkeri by cesium chloride or cesium trifluoroacetate density gradient centrifugation respectively. However, a constant low background of unspecific nucleic acids was detected in all DNA or rRNA gradient fractions, which is important for the interpretation of environmental SIP results. Consequently, quantitative analysis of gradient fractions provides a higher precision and finer resolution for retrieval of isotopically enriched nucleic acids than possible using ethidium bromide or gradient fractionation combined with fingerprinting analyses. This is a prerequisite for the fine-scale tracing of microbial populations metabolizing 13C-labelled compounds in natural ecosystems.

The Impact of Microphysics and Planetary Boundary Layer Physics on Model Simulation of U.S. Deep South Summer Convection

Science.gov (United States)

McCaul, Eugene W., Jr.; Case, Jonathan L.; Zavodsky, Bradley T.; Srikishen, Jayanthi; Medlin, Jeffrey M.; Wood, Lance

2014-01-01

Inspection of output from various configurations of high-resolution, explicit convection forecast models such as the Weather Research and Forecasting (WRF) model indicates significant sensitivity to the choices of model physics pararneterizations employed. Some of the largest apparent sensitivities are related to the specifications of the cloud microphysics and planetary boundary layer physics packages. In addition, these sensitivities appear to be especially pronounced for the weakly-sheared, multicell modes of deep convection characteristic of the Deep South of the United States during the boreal summer. Possible ocean-land sensitivities also argue for further examination of the impacts of using unique ocean-land surface initialization datasets provided by the NASA Short-term Prediction Research and Transition (SPoRn Center to select NOAAlNWS weather forecast offices. To obtain better quantitative understanding of these sensitivities and also to determine the utility of the ocean-land initialization data, we have executed matrices of regional WRF forecasts for selected convective events near Mobile, AL (MOB), and Houston, TX (HGX). The matrices consist of identically initialized WRF 24-h forecasts using any of eight microphysics choices and any of three planetary boWldary layer choices. The resulting 24 simulations performed for each event within either the MOB or HGX regions are then compared to identify the sensitivities of various convective storm metrics to the physics choices. Particular emphasis is placed on sensitivities of precipitation timing, intensity, and coverage, as well as amount and coverage oflightuing activity diagnosed from storm kinematics and graupel in the mixed phase layer. The results confirm impressions gleaned from study of the behavior of variously configured WRF runs contained in the ensembles produced each spring at the Center for the Analysis and Prediction of Storms, but with the benefit of more straightforward control of the

Deep levels in as-grown and Si-implanted In(0.2)Ga(0.8)As-GaAs strained-layer superlattice optical guiding structures

Science.gov (United States)

Dhar, S.; Das, U.; Bhattacharya, P. K.

1986-01-01

Trap levels in about 2-micron In(0.2)Ga(0.8)As(94 A)/GaAs(25 A) strained-layer superlattices, suitable for optical waveguides, have been identified and characterized by deep-level transient spectroscopy and optical deep-level transient spectroscopy measurements. Several dominant electron and hole traps with concentrations of approximately 10 to the 14th/cu cm, and thermal ionization energies Delta-E(T) varying from 0.20 to 0.75 eV have been detected. Except for a 0.20-eV electron trap, which might be present in the In(0.2)Ga(0.8)As well regions, all the other traps have characteristics similar to those identified in molecular-beam epitaxial GaAs. Of these, a 0.42-eV hole trap is believed to originate from Cu impurities, and the others are probably related to native defects. Upon Si implantation and halogen lamp annealing, new deep centers are created. These are electron traps with Delta-E(T) = 0.81 eV and hole traps with Delta-E(T) = 0.46 eV. Traps occurring at room temperature may present limitations for optical devices.

Deep Learning in the Automotive Industry: Applications and Tools

OpenAIRE

Luckow, Andre; Cook, Matthew; Ashcraft, Nathan; Weill, Edwin; Djerekarov, Emil; Vorster, Bennie

2017-01-01

Deep Learning refers to a set of machine learning techniques that utilize neural networks with many hidden layers for tasks, such as image classification, speech recognition, language understanding. Deep learning has been proven to be very effective in these domains and is pervasively used by many Internet services. In this paper, we describe different automotive uses cases for deep learning in particular in the domain of computer vision. We surveys the current state-of-the-art in libraries, ...

Massively Parallel Assimilation of TOGA/TAO and Topex/Poseidon Measurements into a Quasi Isopycnal Ocean General Circulation Model Using an Ensemble Kalman Filter

Science.gov (United States)

Keppenne, Christian L.; Rienecker, Michele; Borovikov, Anna Y.; Suarez, Max

1999-01-01

A massively parallel ensemble Kalman filter (EnKF)is used to assimilate temperature data from the TOGA/TAO array and altimetry from TOPEX/POSEIDON into a Pacific basin version of the NASA Seasonal to Interannual Prediction Project (NSIPP)ls quasi-isopycnal ocean general circulation model. The EnKF is an approximate Kalman filter in which the error-covariance propagation step is modeled by the integration of multiple instances of a numerical model. An estimate of the true error covariances is then inferred from the distribution of the ensemble of model state vectors. This inplementation of the filter takes advantage of the inherent parallelism in the EnKF algorithm by running all the model instances concurrently. The Kalman filter update step also occurs in parallel by having each processor process the observations that occur in the region of physical space for which it is responsible. The massively parallel data assimilation system is validated by withholding some of the data and then quantifying the extent to which the withheld information can be inferred from the assimilation of the remaining data. The distributions of the forecast and analysis error covariances predicted by the ENKF are also examined.

Nonlocal impacts of the Loop Current on cross-slope near-bottom flow in the northeastern Gulf of Mexico

Science.gov (United States)

Nguyen, Thanh-Tam; Morey, Steven L.; Dukhovskoy, Dmitry S.; Chassignet, Eric P.

2015-04-01

Cross-slope near-bottom motions near De Soto Canyon in the northeastern Gulf of Mexico are analyzed from a multidecadal ocean model simulation to characterize upwelling and downwelling, important mechanisms for exchange between the deep ocean and shelf in the vicinity of the 2010 BP Macondo well oil spill. Across the continental slope, large-scale depression and offshore movement of isopycnals (downwelling) occur more frequently when the Loop Current impinges upon the West Florida Shelf slope farther south. Upwelling and onshore movement of isopycnals occurs with roughly the same likelihood regardless of Loop Current impingement on the slope. The remote influence of Loop Current on the De Soto Canyon region downwelling is a consequence of a high-pressure anomaly that extends along the continental slope emanating from the location of Loop Current impact.

Deep Transfer Metric Learning.

Science.gov (United States)

Junlin Hu; Jiwen Lu; Yap-Peng Tan; Jie Zhou

2016-12-01

Conventional metric learning methods usually assume that the training and test samples are captured in similar scenarios so that their distributions are assumed to be the same. This assumption does not hold in many real visual recognition applications, especially when samples are captured across different data sets. In this paper, we propose a new deep transfer metric learning (DTML) method to learn a set of hierarchical nonlinear transformations for cross-domain visual recognition by transferring discriminative knowledge from the labeled source domain to the unlabeled target domain. Specifically, our DTML learns a deep metric network by maximizing the inter-class variations and minimizing the intra-class variations, and minimizing the distribution divergence between the source domain and the target domain at the top layer of the network. To better exploit the discriminative information from the source domain, we further develop a deeply supervised transfer metric learning (DSTML) method by including an additional objective on DTML, where the output of both the hidden layers and the top layer are optimized jointly. To preserve the local manifold of input data points in the metric space, we present two new methods, DTML with autoencoder regularization and DSTML with autoencoder regularization. Experimental results on face verification, person re-identification, and handwritten digit recognition validate the effectiveness of the proposed methods.

Influence of a deep-level-defect band formed in a heavily Mg-doped GaN contact layer on the Ni/Au contact to p-GaN

International Nuclear Information System (INIS)

Li Xiao-Jing; Zhao De-Gang; Jiang De-Sheng; Chen Ping; Zhu Jian-Jun; Liu Zong-Shun; Yang Jing; He Xiao-Guang; Yang Hui; Zhang Li-Qun; Zhang Shu-Ming; Le Ling-Cong; Liu Jian-Ping

2015-01-01

The influence of a deep-level-defect (DLD) band formed in a heavily Mg-doped GaN contact layer on the performance of Ni/Au contact to p-GaN is investigated. The thin heavily Mg-doped GaN (p ++ -GaN) contact layer with DLD band can effectively improve the performance of Ni/Au ohmic contact to p-GaN. The temperature-dependent I–V measurement shows that the variable-range hopping (VRH) transportation through the DLD band plays a dominant role in the ohmic contact. The thickness and Mg/Ga flow ratio of p ++ -GaN contact layer have a significant effect on ohmic contact by controlling the Mg impurity doping and the formation of a proper DLD band. When the thickness of the p ++ -GaN contact layer is 25 nm thick and the Mg/Ga flow rate ratio is 10.29%, an ohmic contact with low specific contact resistivity of 6.97× 10 −4 Ω·cm 2 is achieved. (paper)

Electronic structure properties of deep defects in hBN

Science.gov (United States)

Dev, Pratibha; Prdm Collaboration

In recent years, the search for room-temperature solid-state qubit (quantum bit) candidates has revived interest in the study of deep-defect centers in semiconductors. The charged NV-center in diamond is the best known amongst these defects. However, as a host material, diamond poses several challenges and so, increasingly, there is an interest in exploring deep defects in alternative semiconductors such as hBN. The layered structure of hBN makes it a scalable platform for quantum applications, as there is a greater potential for controlling the location of the deep defect in the 2D-matrix through careful experiments. Using density functional theory-based methods, we have studied the electronic and structural properties of several deep defects in hBN. Native defects within hBN layers are shown to have high spin ground states that should survive even at room temperature, making them interesting solid-state qubit candidates in a 2D matrix. Partnership for Reduced Dimensional Material (PRDM) is part of the NSF sponsored Partnerships for Research and Education in Materials (PREM).

Processing of chromatic information in a deep convolutional neural network.

Science.gov (United States)

Flachot, Alban; Gegenfurtner, Karl R

2018-04-01

Deep convolutional neural networks are a class of machine-learning algorithms capable of solving non-trivial tasks, such as object recognition, with human-like performance. Little is known about the exact computations that deep neural networks learn, and to what extent these computations are similar to the ones performed by the primate brain. Here, we investigate how color information is processed in the different layers of the AlexNet deep neural network, originally trained on object classification of over 1.2M images of objects in their natural contexts. We found that the color-responsive units in the first layer of AlexNet learned linear features and were broadly tuned to two directions in color space, analogously to what is known of color responsive cells in the primate thalamus. Moreover, these directions are decorrelated and lead to statistically efficient representations, similar to the cardinal directions of the second-stage color mechanisms in primates. We also found, in analogy to the early stages of the primate visual system, that chromatic and achromatic information were segregated in the early layers of the network. Units in the higher layers of AlexNet exhibit on average a lower responsivity for color than units at earlier stages.

Layer-specific morphological and molecular differences in neocortical astrocytes and their dependence on neuronal layers.

Science.gov (United States)

Lanjakornsiripan, Darin; Pior, Baek-Jun; Kawaguchi, Daichi; Furutachi, Shohei; Tahara, Tomoaki; Katsuyama, Yu; Suzuki, Yutaka; Fukazawa, Yugo; Gotoh, Yukiko

2018-04-24

Non-pial neocortical astrocytes have historically been thought to comprise largely a nondiverse population of protoplasmic astrocytes. Here we show that astrocytes of the mouse somatosensory cortex manifest layer-specific morphological and molecular differences. Two- and three-dimensional observations revealed that astrocytes in the different layers possess distinct morphologies as reflected by differences in cell orientation, territorial volume, and arborization. The extent of ensheathment of synaptic clefts by astrocytes in layer II/III was greater than that by those in layer VI. Moreover, differences in gene expression were observed between upper-layer and deep-layer astrocytes. Importantly, layer-specific differences in astrocyte properties were abrogated in reeler and Dab1 conditional knockout mice, in which neuronal layers are disturbed, suggesting that neuronal layers are a prerequisite for the observed morphological and molecular differences of neocortical astrocytes. This study thus demonstrates the existence of layer-specific interactions between neurons and astrocytes, which may underlie their layer-specific functions.

Shakeout: A New Approach to Regularized Deep Neural Network Training.

Science.gov (United States)

Kang, Guoliang; Li, Jun; Tao, Dacheng

2018-05-01

Recent years have witnessed the success of deep neural networks in dealing with a plenty of practical problems. Dropout has played an essential role in many successful deep neural networks, by inducing regularization in the model training. In this paper, we present a new regularized training approach: Shakeout. Instead of randomly discarding units as Dropout does at the training stage, Shakeout randomly chooses to enhance or reverse each unit's contribution to the next layer. This minor modification of Dropout has the statistical trait: the regularizer induced by Shakeout adaptively combines , and regularization terms. Our classification experiments with representative deep architectures on image datasets MNIST, CIFAR-10 and ImageNet show that Shakeout deals with over-fitting effectively and outperforms Dropout. We empirically demonstrate that Shakeout leads to sparser weights under both unsupervised and supervised settings. Shakeout also leads to the grouping effect of the input units in a layer. Considering the weights in reflecting the importance of connections, Shakeout is superior to Dropout, which is valuable for the deep model compression. Moreover, we demonstrate that Shakeout can effectively reduce the instability of the training process of the deep architecture.

Deep Correlated Holistic Metric Learning for Sketch-Based 3D Shape Retrieval.

Science.gov (United States)

Dai, Guoxian; Xie, Jin; Fang, Yi

2018-07-01

How to effectively retrieve desired 3D models with simple queries is a long-standing problem in computer vision community. The model-based approach is quite straightforward but nontrivial, since people could not always have the desired 3D query model available by side. Recently, large amounts of wide-screen electronic devices are prevail in our daily lives, which makes the sketch-based 3D shape retrieval a promising candidate due to its simpleness and efficiency. The main challenge of sketch-based approach is the huge modality gap between sketch and 3D shape. In this paper, we proposed a novel deep correlated holistic metric learning (DCHML) method to mitigate the discrepancy between sketch and 3D shape domains. The proposed DCHML trains two distinct deep neural networks (one for each domain) jointly, which learns two deep nonlinear transformations to map features from both domains into a new feature space. The proposed loss, including discriminative loss and correlation loss, aims to increase the discrimination of features within each domain as well as the correlation between different domains. In the new feature space, the discriminative loss minimizes the intra-class distance of the deep transformed features and maximizes the inter-class distance of the deep transformed features to a large margin within each domain, while the correlation loss focused on mitigating the distribution discrepancy across different domains. Different from existing deep metric learning methods only with loss at the output layer, our proposed DCHML is trained with loss at both hidden layer and output layer to further improve the performance by encouraging features in the hidden layer also with desired properties. Our proposed method is evaluated on three benchmarks, including 3D Shape Retrieval Contest 2013, 2014, and 2016 benchmarks, and the experimental results demonstrate the superiority of our proposed method over the state-of-the-art methods.

Deep neural mapping support vector machines.

Science.gov (United States)

Li, Yujian; Zhang, Ting

2017-09-01

The choice of kernel has an important effect on the performance of a support vector machine (SVM). The effect could be reduced by NEUROSVM, an architecture using multilayer perceptron for feature extraction and SVM for classification. In binary classification, a general linear kernel NEUROSVM can be theoretically simplified as an input layer, many hidden layers, and an SVM output layer. As a feature extractor, the sub-network composed of the input and hidden layers is first trained together with a virtual ordinary output layer by backpropagation, then with the output of its last hidden layer taken as input of the SVM classifier for further training separately. By taking the sub-network as a kernel mapping from the original input space into a feature space, we present a novel model, called deep neural mapping support vector machine (DNMSVM), from the viewpoint of deep learning. This model is also a new and general kernel learning method, where the kernel mapping is indeed an explicit function expressed as a sub-network, different from an implicit function induced by a kernel function traditionally. Moreover, we exploit a two-stage procedure of contrastive divergence learning and gradient descent for DNMSVM to jointly training an adaptive kernel mapping instead of a kernel function, without requirement of kernel tricks. As a whole of the sub-network and the SVM classifier, the joint training of DNMSVM is done by using gradient descent to optimize the objective function with the sub-network layer-wise pre-trained via contrastive divergence learning of restricted Boltzmann machines. Compared to the separate training of NEUROSVM, the joint training is a new algorithm for DNMSVM to have advantages over NEUROSVM. Experimental results show that DNMSVM can outperform NEUROSVM and RBFSVM (i.e., SVM with the kernel of radial basis function), demonstrating its effectiveness. Copyright © 2017 Elsevier Ltd. All rights reserved.

Spectroscopy of Deep Traps in Cu2S-CdS Junction Structures

Directory of Open Access Journals (Sweden)

Eugenijus Gaubas

2012-12-01

Full Text Available Cu2S-CdS junctions of the polycrystalline material layers have been examined by combining the capacitance deep level transient spectroscopy technique together with white LED light additional illumination (C-DLTS-WL and the photo-ionization spectroscopy (PIS implemented by the photocurrent probing. Three types of junction structures, separated by using the barrier capacitance characteristics of the junctions and correlated with XRD distinguished precipitates of the polycrystalline layers, exhibit different deep trap spectra within CdS substrates.

Applications of Deep Learning in Biomedicine.

Science.gov (United States)

Mamoshina, Polina; Vieira, Armando; Putin, Evgeny; Zhavoronkov, Alex

2016-05-02

Increases in throughput and installed base of biomedical research equipment led to a massive accumulation of -omics data known to be highly variable, high-dimensional, and sourced from multiple often incompatible data platforms. While this data may be useful for biomarker identification and drug discovery, the bulk of it remains underutilized. Deep neural networks (DNNs) are efficient algorithms based on the use of compositional layers of neurons, with advantages well matched to the challenges -omics data presents. While achieving state-of-the-art results and even surpassing human accuracy in many challenging tasks, the adoption of deep learning in biomedicine has been comparatively slow. Here, we discuss key features of deep learning that may give this approach an edge over other machine learning methods. We then consider limitations and review a number of applications of deep learning in biomedical studies demonstrating proof of concept and practical utility.

An improved advertising CTR prediction approach based on the fuzzy deep neural network.

Science.gov (United States)

Jiang, Zilong; Gao, Shu; Li, Mingjiang

2018-01-01

Combining a deep neural network with fuzzy theory, this paper proposes an advertising click-through rate (CTR) prediction approach based on a fuzzy deep neural network (FDNN). In this approach, fuzzy Gaussian-Bernoulli restricted Boltzmann machine (FGBRBM) is first applied to input raw data from advertising datasets. Next, fuzzy restricted Boltzmann machine (FRBM) is used to construct the fuzzy deep belief network (FDBN) with the unsupervised method layer by layer. Finally, fuzzy logistic regression (FLR) is utilized for modeling the CTR. The experimental results show that the proposed FDNN model outperforms several baseline models in terms of both data representation capability and robustness in advertising click log datasets with noise.

Deep-level transient spectroscopy of low-energy ion-irradiated silicon

DEFF Research Database (Denmark)

Kolkovsky, Vladimir; Privitera, V.; Nylandsted Larsen, Arne

2009-01-01

 During electron-gun deposition of metal layers on semiconductors, the semiconductor is bombarded with low-energy metal ions creating defects in the outermost surface layer. For many years, it has been a puzzle why deep-level transient spectroscopy spectra of the as-deposited, electron-gun evapor...

Crude oil treatment leads to shift of bacterial communities in soils from the deep active layer and upper permafrost along the China-Russia Crude Oil Pipeline route.

Science.gov (United States)

Yang, Sizhong; Wen, Xi; Zhao, Liang; Shi, Yulan; Jin, Huijun

2014-01-01

The buried China-Russia Crude Oil Pipeline (CRCOP) across the permafrost-associated cold ecosystem in northeastern China carries a risk of contamination to the deep active layers and upper permafrost in case of accidental rupture of the embedded pipeline or migration of oil spills. As many soil microbes are capable of degrading petroleum, knowledge about the intrinsic degraders and the microbial dynamics in the deep subsurface could extend our understanding of the application of in-situ bioremediation. In this study, an experiment was conducted to investigate the bacterial communities in response to simulated contamination to deep soil samples by using 454 pyrosequencing amplicons. The result showed that bacterial diversity was reduced after 8-weeks contamination. A shift in bacterial community composition was apparent in crude oil-amended soils with Proteobacteria (esp. α-subdivision) being the dominant phylum, together with Actinobacteria and Firmicutes. The contamination led to enrichment of indigenous bacterial taxa like Novosphingobium, Sphingobium, Caulobacter, Phenylobacterium, Alicylobacillus and Arthrobacter, which are generally capable of degrading polycyclic aromatic hydrocarbons (PAHs). The community shift highlighted the resilience of PAH degraders and their potential for in-situ degradation of crude oil under favorable conditions in the deep soils.

Crude oil treatment leads to shift of bacterial communities in soils from the deep active layer and upper permafrost along the China-Russia Crude Oil Pipeline route.

Directory of Open Access Journals (Sweden)

Sizhong Yang

Full Text Available The buried China-Russia Crude Oil Pipeline (CRCOP across the permafrost-associated cold ecosystem in northeastern China carries a risk of contamination to the deep active layers and upper permafrost in case of accidental rupture of the embedded pipeline or migration of oil spills. As many soil microbes are capable of degrading petroleum, knowledge about the intrinsic degraders and the microbial dynamics in the deep subsurface could extend our understanding of the application of in-situ bioremediation. In this study, an experiment was conducted to investigate the bacterial communities in response to simulated contamination to deep soil samples by using 454 pyrosequencing amplicons. The result showed that bacterial diversity was reduced after 8-weeks contamination. A shift in bacterial community composition was apparent in crude oil-amended soils with Proteobacteria (esp. α-subdivision being the dominant phylum, together with Actinobacteria and Firmicutes. The contamination led to enrichment of indigenous bacterial taxa like Novosphingobium, Sphingobium, Caulobacter, Phenylobacterium, Alicylobacillus and Arthrobacter, which are generally capable of degrading polycyclic aromatic hydrocarbons (PAHs. The community shift highlighted the resilience of PAH degraders and their potential for in-situ degradation of crude oil under favorable conditions in the deep soils.

Crude Oil Treatment Leads to Shift of Bacterial Communities in Soils from the Deep Active Layer and Upper Permafrost along the China-Russia Crude Oil Pipeline Route

Science.gov (United States)

Yang, Sizhong; Wen, Xi; Zhao, Liang; Shi, Yulan; Jin, Huijun

2014-01-01

The buried China-Russia Crude Oil Pipeline (CRCOP) across the permafrost-associated cold ecosystem in northeastern China carries a risk of contamination to the deep active layers and upper permafrost in case of accidental rupture of the embedded pipeline or migration of oil spills. As many soil microbes are capable of degrading petroleum, knowledge about the intrinsic degraders and the microbial dynamics in the deep subsurface could extend our understanding of the application of in-situ bioremediation. In this study, an experiment was conducted to investigate the bacterial communities in response to simulated contamination to deep soil samples by using 454 pyrosequencing amplicons. The result showed that bacterial diversity was reduced after 8-weeks contamination. A shift in bacterial community composition was apparent in crude oil-amended soils with Proteobacteria (esp. α-subdivision) being the dominant phylum, together with Actinobacteria and Firmicutes. The contamination led to enrichment of indigenous bacterial taxa like Novosphingobium, Sphingobium, Caulobacter, Phenylobacterium, Alicylobacillus and Arthrobacter, which are generally capable of degrading polycyclic aromatic hydrocarbons (PAHs). The community shift highlighted the resilience of PAH degraders and their potential for in-situ degradation of crude oil under favorable conditions in the deep soils. PMID:24794099

Low temperature sacrificial wafer bonding for planarization after very deep etching

NARCIS (Netherlands)

Spiering, V.L.; Spiering, V.L.; Berenschot, Johan W.; Elwenspoek, Michael Curt; Fluitman, J.H.J.

1994-01-01

A new technique, at temperatures of 150°C or 450°C, that provides planarization after a very deep etching step in silicon is presented. Resist spinning and layer patterning as well as realization of bridges or cantilevers across deep holes becomes possible. The sacrificial wafer bonding technique

Unexpectedly high soil organic carbon stocks under impervious surfaces contributed by urban deep cultural layers

Science.gov (United States)

Bae, J.; Ryu, Y.

2017-12-01

The expansion of urban artificial structures has altered the spatial distribution of soil organic carbon (SOC) stocks. The majority of the urban soil studies within the land-cover types, however, focused on top soils despite the potential of deep soils to store large amounts of SOC. Here, we investigate vertical distribution of SOC stocks in both impervious surfaces (n = 11) and adjacent green spaces (n = 8) to a depth of 4 m with in an apartment complex area, Seoul, Republic of Korea. We found that more than six times differences in SOC stocks were observed at 0-1 m depth between the impervious surfaces (1.90 kgC m-2) and the green spaces (12.03 kgC m-2), but no significant differences appeared when comparing them at the depth of 0-4 m. We found "cultural layers" with the largest SOC stocks at 1-2 m depth in the impervious surfaces (15.85 kgC m-2) and 2-3 m depths in urban green spaces (12.52 kgC m-2). Thus, the proportions of SOC stocks at the 0-1 m depth to the total of 0-4 m depth were 6.83% in impervious surfaces and 32.15% in urban green spaces, respectively. The 13C and 15N stable isotope data with historical aerial photographs revealed that the cropland which existed before 1978 formed the SOC in the cultural layers. Our results highlight that impervious surface could hold large amount of SOC stock which has been overlooked in urban carbon cycles. We believe this finding will help city planners and policy makers to develop carbon management programs better towards sustainable urban ecosystems.

A simple model of the effect of ocean ventilation on ocean heat uptake

Science.gov (United States)

Nadiga, Balu; Urban, Nathan

2017-11-01

Transport of water from the surface mixed layer into the ocean interior is achieved, in large part, by the process of ventilation-a process associated with outcropping isopycnals. Starting from such a configuration of outcropping isopycnals, we derive a simple model of the effect of ventilation on ocean uptake of anomalous radiative forcing. This model can be seen as an improvement of the popular anomaly-diffusing class of energy balance models (AD-EBM) that are routinely employed to analyze and emulate the warming response of both observed and simulated Earth system. We demonstrate that neither multi-layer, nor continuous-diffusion AD-EBM variants can properly represent both surface-warming and the vertical distribution of ocean heat uptake. The new model overcomes this deficiency. The simplicity of the models notwithstanding, the analysis presented and the necessity of the modification is indicative of the role played by processes related to the down-welling branch of global ocean circulation in shaping the vertical distribution of ocean heat uptake.

Atomic Layer Deposition of Chemical Passivation Layers and High Performance Anti-Reflection Coatings on Back-Illuminated Detectors

Science.gov (United States)

Hoenk, Michael E. (Inventor); Greer, Frank (Inventor); Nikzad, Shouleh (Inventor)

2014-01-01

A back-illuminated silicon photodetector has a layer of Al2O3 deposited on a silicon oxide surface that receives electromagnetic radiation to be detected. The Al2O3 layer has an antireflection coating deposited thereon. The Al2O3 layer provides a chemically resistant separation layer between the silicon oxide surface and the antireflection coating. The Al2O3 layer is thin enough that it is optically innocuous. Under deep ultraviolet radiation, the silicon oxide layer and the antireflection coating do not interact chemically. In one embodiment, the silicon photodetector has a delta-doped layer near (within a few nanometers of) the silicon oxide surface. The Al2O3 layer is expected to provide similar protection for doped layers fabricated using other methods, such as MBE, ion implantation and CVD deposition.

Complex approach mechanical properties and formability assessment of selected deep-drawing steels

Directory of Open Access Journals (Sweden)

J. Štaba

2009-07-01

Full Text Available The paper analyses the properties of deep-drawing sheets of three grades (Re = 320 to 475 MPa, surface-treated with hot-dip galvanizing, made of microalloyed steels. Deformation properties are assessed using tensile tests, technological Erichsen or cupping tests. These characteristics, as well as the behaviour of the surface layer, are also investigated under dynamic conditions (modified Erichsen test using a drop tester, or using flat bending fatigue tests. Using microscopic analysis the deformation properties of the surface layer are evaluated. The results show the compactness of the surface layer, high deformation characteristics, as well as fatigue properties of the investigated deep-drawing materials, suitable for application in the automotive industry.

The deep lymphatic anatomy of the hand.

Science.gov (United States)

Ma, Chuan-Xiang; Pan, Wei-Ren; Liu, Zhi-An; Zeng, Fan-Qiang; Qiu, Zhi-Qiang

2018-04-03

The deep lymphatic anatomy of the hand still remains the least described in medical literature. Eight hands were harvested from four nonembalmed human cadavers amputated above the wrist. A small amount of 6% hydrogen peroxide was employed to detect the lymphatic vessels around the superficial and deep palmar vascular arches, in webs from the index to little fingers, the thenar and hypothenar areas. A 30-gauge needle was inserted into the vessels and injected with a barium sulphate compound. Each specimen was dissected, photographed and radiographed to demonstrate deep lymphatic distribution of the hand. Five groups of deep collecting lymph vessels were found in the hand: superficial palmar arch lymph vessel (SPALV); deep palmar arch lymph vessel (DPALV); thenar lymph vessel (TLV); hypothenar lymph vessel (HTLV); deep finger web lymph vessel (DFWLV). Each group of vessels drained in different directions first, then all turned and ran towards the wrist in different layers. The deep lymphatic drainage of the hand has been presented. The results will provide an anatomical basis for clinical management, educational reference and scientific research. Copyright © 2018 Elsevier GmbH. All rights reserved.

Statistical-Mechanical Analysis of Pre-training and Fine Tuning in Deep Learning

Science.gov (United States)

Ohzeki, Masayuki

2015-03-01

In this paper, we present a statistical-mechanical analysis of deep learning. We elucidate some of the essential components of deep learning — pre-training by unsupervised learning and fine tuning by supervised learning. We formulate the extraction of features from the training data as a margin criterion in a high-dimensional feature-vector space. The self-organized classifier is then supplied with small amounts of labelled data, as in deep learning. Although we employ a simple single-layer perceptron model, rather than directly analyzing a multi-layer neural network, we find a nontrivial phase transition that is dependent on the number of unlabelled data in the generalization error of the resultant classifier. In this sense, we evaluate the efficacy of the unsupervised learning component of deep learning. The analysis is performed by the replica method, which is a sophisticated tool in statistical mechanics. We validate our result in the manner of deep learning, using a simple iterative algorithm to learn the weight vector on the basis of belief propagation.

Deep Learning- and Transfer Learning-Based Super Resolution Reconstruction from Single Medical Image

Directory of Open Access Journals (Sweden)

YiNan Zhang

2017-01-01

Full Text Available Medical images play an important role in medical diagnosis and research. In this paper, a transfer learning- and deep learning-based super resolution reconstruction method is introduced. The proposed method contains one bicubic interpolation template layer and two convolutional layers. The bicubic interpolation template layer is prefixed by mathematics deduction, and two convolutional layers learn from training samples. For saving training medical images, a SIFT feature-based transfer learning method is proposed. Not only can medical images be used to train the proposed method, but also other types of images can be added into training dataset selectively. In empirical experiments, results of eight distinctive medical images show improvement of image quality and time reduction. Further, the proposed method also produces slightly sharper edges than other deep learning approaches in less time and it is projected that the hybrid architecture of prefixed template layer and unfixed hidden layers has potentials in other applications.

Breast cancer molecular subtype classification using deep features: preliminary results

Science.gov (United States)

Zhu, Zhe; Albadawy, Ehab; Saha, Ashirbani; Zhang, Jun; Harowicz, Michael R.; Mazurowski, Maciej A.

2018-02-01

Radiogenomics is a field of investigation that attempts to examine the relationship between imaging characteris- tics of cancerous lesions and their genomic composition. This could offer a noninvasive alternative to establishing genomic characteristics of tumors and aid cancer treatment planning. While deep learning has shown its supe- riority in many detection and classification tasks, breast cancer radiogenomic data suffers from a very limited number of training examples, which renders the training of the neural network for this problem directly and with no pretraining a very difficult task. In this study, we investigated an alternative deep learning approach referred to as deep features or off-the-shelf network approach to classify breast cancer molecular subtypes using breast dynamic contrast enhanced MRIs. We used the feature maps of different convolution layers and fully connected layers as features and trained support vector machines using these features for prediction. For the feature maps that have multiple layers, max-pooling was performed along each channel. We focused on distinguishing the Luminal A subtype from other subtypes. To evaluate the models, 10 fold cross-validation was performed and the final AUC was obtained by averaging the performance of all the folds. The highest average AUC obtained was 0.64 (0.95 CI: 0.57-0.71), using the feature maps of the last fully connected layer. This indicates the promise of using this approach to predict the breast cancer molecular subtypes. Since the best performance appears in the last fully connected layer, it also implies that breast cancer molecular subtypes may relate to high level image features

Deep primary production in coastal pelagic systems

DEFF Research Database (Denmark)

Lyngsgaard, Maren Moltke; Richardson, Katherine; Markager, Stiig

2014-01-01

produced. The primary production (PP) occurring below the surface layer, i.e. in the pycnocline-bottom layer (PBL), is shown to contribute significantly to total PP. Oxygen concentrations in the PBL are shown to correlate significantly with the deep primary production (DPP) as well as with salinity...... that eutrophication effects may include changes in the structure of planktonic food webs and element cycling in the water column, both brought about through an altered vertical distribution of PP....

Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status.

Science.gov (United States)

Korfiatis, Panagiotis; Kline, Timothy L; Lachance, Daniel H; Parney, Ian F; Buckner, Jan C; Erickson, Bradley J

2017-10-01

Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90% (+/- 3.92%) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72% (+/- 13.61%) while ResNet18 (18 layers) accuracy was 76.75% (+/- 20.67%). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p deep neural architectures can be used to predict molecular biomarkers from routine medical images.

Distinctive Microbial Community Structure in Highly Stratified Deep-Sea Brine Water Columns

KAUST Repository

Bougouffa, Salim; Yang, J. K.; Lee, O. O.; Wang, Y.; Batang, Zenon B.; Al-Suwailem, Abdulaziz M.; Qian, P. Y.

2013-01-01

Atlantis II and Discovery are two hydrothermal and hypersaline deep-sea pools in the Red Sea rift that are characterized by strong thermohalo-stratification and temperatures steadily peaking near the bottom. We conducted comprehensive vertical profiling of the microbial populations in both pools and highlighted the influential environmental factors. Pyrosequencing of the 16S rRNA genes revealed shifts in community structures vis-à-vis depth. High diversity and low abundance were features of the deepest convective layers despite the low cell density. Surprisingly, the brine interfaces had significantly higher cell counts than the overlying deep-sea water, yet they were lowest in diversity. Vertical stratification of the bacterial populations was apparent as we moved from the Alphaproteobacteria-dominated deep sea to the Planctomycetaceae- or Deferribacteres-dominated interfaces to the Gammaproteobacteria-dominated brine layers. Archaeal marine group I was dominant in the deep-sea water and interfaces, while several euryarchaeotic groups increased in the brine. Across sites, microbial phylotypes and abundances varied substantially in the brine interface of Discovery compared with Atlantis II, despite the near-identical populations in the overlying deep-sea waters. The lowest convective layers harbored interestingly similar microbial communities, even though temperature and heavy metal concentrations were very different. Multivariate analysis indicated that temperature and salinity were the major influences shaping the communities. The harsh conditions and the low-abundance phylotypes could explain the observed correlation in the brine pools.

Distinctive Microbial Community Structure in Highly Stratified Deep-Sea Brine Water Columns

KAUST Repository

Bougouffa, Salim

2013-03-29

Atlantis II and Discovery are two hydrothermal and hypersaline deep-sea pools in the Red Sea rift that are characterized by strong thermohalo-stratification and temperatures steadily peaking near the bottom. We conducted comprehensive vertical profiling of the microbial populations in both pools and highlighted the influential environmental factors. Pyrosequencing of the 16S rRNA genes revealed shifts in community structures vis-à-vis depth. High diversity and low abundance were features of the deepest convective layers despite the low cell density. Surprisingly, the brine interfaces had significantly higher cell counts than the overlying deep-sea water, yet they were lowest in diversity. Vertical stratification of the bacterial populations was apparent as we moved from the Alphaproteobacteria-dominated deep sea to the Planctomycetaceae- or Deferribacteres-dominated interfaces to the Gammaproteobacteria-dominated brine layers. Archaeal marine group I was dominant in the deep-sea water and interfaces, while several euryarchaeotic groups increased in the brine. Across sites, microbial phylotypes and abundances varied substantially in the brine interface of Discovery compared with Atlantis II, despite the near-identical populations in the overlying deep-sea waters. The lowest convective layers harbored interestingly similar microbial communities, even though temperature and heavy metal concentrations were very different. Multivariate analysis indicated that temperature and salinity were the major influences shaping the communities. The harsh conditions and the low-abundance phylotypes could explain the observed correlation in the brine pools.

Distribution of phytoplankton groups within the deep chlorophyll maximum

KAUST Repository

Latasa, Mikel

2016-11-01

The fine vertical distribution of phytoplankton groups within the deep chlorophyll maximum (DCM) was studied in the NE Atlantic during summer stratification. A simple but unconventional sampling strategy allowed examining the vertical structure with ca. 2 m resolution. The distribution of Prochlorococcus, Synechococcus, chlorophytes, pelagophytes, small prymnesiophytes, coccolithophores, diatoms, and dinoflagellates was investigated with a combination of pigment-markers, flow cytometry and optical and FISH microscopy. All groups presented minimum abundances at the surface and a maximum in the DCM layer. The cell distribution was not vertically symmetrical around the DCM peak and cells tended to accumulate in the upper part of the DCM layer. The more symmetrical distribution of chlorophyll than cells around the DCM peak was due to the increase of pigment per cell with depth. We found a vertical alignment of phytoplankton groups within the DCM layer indicating preferences for different ecological niches in a layer with strong gradients of light and nutrients. Prochlorococcus occupied the shallowest and diatoms the deepest layers. Dinoflagellates, Synechococcus and small prymnesiophytes preferred shallow DCM layers, and coccolithophores, chlorophytes and pelagophytes showed a preference for deep layers. Cell size within groups changed with depth in a pattern related to their mean size: the cell volume of the smallest group increased the most with depth while the cell volume of the largest group decreased the most. The vertical alignment of phytoplankton groups confirms that the DCM is not a homogeneous entity and indicates groups’ preferences for different ecological niches within this layer.

Two-layer anti-reflection strategies for implant applications

Science.gov (United States)

Guerrero, Douglas J.; Smith, Tamara; Kato, Masakazu; Kimura, Shigeo; Enomoto, Tomoyuki

2006-03-01

A two-layer bottom anti-reflective coating (BARC) concept in which a layer that develops slowly is coated on top of a bottom layer that develops more rapidly was demonstrated. Development rate control was achieved by selection of crosslinker amount and BARC curing conditions. A single-layer BARC was compared with the two-layer BARC concept. The single-layer BARC does not clear out of 200-nm deep vias. When the slower developing single-layer BARC was coated on top of the faster developing layer, the vias were cleared. Lithographic evaluation of the two-layer BARC concept shows the same resolution advantages as the single-layer system. Planarization properties of a two-layer BARC system are better than for a single-layer system, when comparing the same total nominal thicknesses.

A new approach to develop computer-aided diagnosis scheme of breast mass classification using deep learning technology.

Science.gov (United States)

Qiu, Yuchen; Yan, Shiju; Gundreddy, Rohith Reddy; Wang, Yunzhi; Cheng, Samuel; Liu, Hong; Zheng, Bin

2017-01-01

To develop and test a deep learning based computer-aided diagnosis (CAD) scheme of mammograms for classifying between malignant and benign masses. An image dataset involving 560 regions of interest (ROIs) extracted from digital mammograms was used. After down-sampling each ROI from 512×512 to 64×64 pixel size, we applied an 8 layer deep learning network that involves 3 pairs of convolution-max-pooling layers for automatic feature extraction and a multiple layer perceptron (MLP) classifier for feature categorization to process ROIs. The 3 pairs of convolution layers contain 20, 10, and 5 feature maps, respectively. Each convolution layer is connected with a max-pooling layer to improve the feature robustness. The output of the sixth layer is fully connected with a MLP classifier, which is composed of one hidden layer and one logistic regression layer. The network then generates a classification score to predict the likelihood of ROI depicting a malignant mass. A four-fold cross validation method was applied to train and test this deep learning network. The results revealed that this CAD scheme yields an area under the receiver operation characteristic curve (AUC) of 0.696±0.044, 0.802±0.037, 0.836±0.036, and 0.822±0.035 for fold 1 to 4 testing datasets, respectively. The overall AUC of the entire dataset is 0.790±0.019. This study demonstrates the feasibility of applying a deep learning based CAD scheme to classify between malignant and benign breast masses without a lesion segmentation, image feature computation and selection process.

A New Approach to Develop Computer-aided Diagnosis Scheme of Breast Mass Classification Using Deep Learning Technology

Science.gov (United States)

Qiu, Yuchen; Yan, Shiju; Gundreddy, Rohith Reddy; Wang, Yunzhi; Cheng, Samuel; Liu, Hong; Zheng, Bin

2017-01-01

PURPOSE To develop and test a deep learning based computer-aided diagnosis (CAD) scheme of mammograms for classifying between malignant and benign masses. METHODS An image dataset involving 560 regions of interest (ROIs) extracted from digital mammograms was used. After down-sampling each ROI from 512×512 to 64×64 pixel size, we applied an 8 layer deep learning network that involves 3 pairs of convolution-max-pooling layers for automatic feature extraction and a multiple layer perceptron (MLP) classifier for feature categorization to process ROIs. The 3 pairs of convolution layers contain 20, 10, and 5 feature maps, respectively. Each convolution layer is connected with a max-pooling layer to improve the feature robustness. The output of the sixth layer is fully connected with a MLP classifier, which is composed of one hidden layer and one logistic regression layer. The network then generates a classification score to predict the likelihood of ROI depicting a malignant mass. A four-fold cross validation method was applied to train and test this deep learning network. RESULTS The results revealed that this CAD scheme yields an area under the receiver operation characteristic curve (AUC) of 0.696±0.044, 0.802±0.037, 0.836±0.036, and 0.822±0.035 for fold 1 to 4 testing datasets, respectively. The overall AUC of the entire dataset is 0.790±0.019. CONCLUSIONS This study demonstrates the feasibility of applying a deep learning based CAD scheme to classify between malignant and benign breast masses without a lesion segmentation, image feature computation and selection process. PMID:28436410

Exploring orange peel treatment with deep eutectic solvents and diluted organic acids

NARCIS (Netherlands)

van den Bruinhorst, A.; Kouris, P.; Timmer, J.M.K.; de Croon, M.H.J.M.; Kroon, M.C.

2016-01-01

The disintegration of orange peel waste in deep eutectic solvents and diluted organic acids is presented in this work. The albedo and flavedo layers of the peel were studied separately, showing faster disintegration of the latter. Addition of water to the deep eutectic solvents lowered the amount of

A reason of fast and deep fading of centimeter wave

International Nuclear Information System (INIS)

Danzan, D.; Damdinsuren, E.; Hiamjav, J.; Chuluunbaatar, Ch.; Battulga, S.

1992-01-01

First discovered experimentally exactly correlation between of appearance and of disappearance of optical mirage and fast and deep fading of horizontal polarization of centimeter wave. Proved the interference of the straight and reflected rays from the thin layer of air in mirage a reason of this fading. The physical parameters data of the layer of mirage: change of dielectric permeability and n/ h gradient of refraction index of air in this layer are been showed

Application of deep convolutional neural networks for ocean front recognition

Science.gov (United States)

Lima, Estanislau; Sun, Xin; Yang, Yuting; Dong, Junyu

2017-10-01

Ocean fronts have been a subject of study for many years, a variety of methods and algorithms have been proposed to address the problem of ocean fronts. However, all these existing ocean front recognition methods are built upon human expertise in defining the front based on subjective thresholds of relevant physical variables. This paper proposes a deep learning approach for ocean front recognition that is able to automatically recognize the front. We first investigated four existing deep architectures, i.e., AlexNet, CaffeNet, GoogLeNet, and VGGNet, for the ocean front recognition task using remote sensing (RS) data. We then propose a deep network with fewer layers compared to existing architecture for the front recognition task. This network has a total of five learnable layers. In addition, we extended the proposed network to recognize and classify the front into strong and weak ones. We evaluated and analyzed the proposed network with two strategies of exploiting the deep model: full-training and fine-tuning. Experiments are conducted on three different RS image datasets, which have different properties. Experimental results show that our model can produce accurate recognition results.

Stellar Atmospheric Parameterization Based on Deep Learning

Science.gov (United States)

Pan, Ru-yang; Li, Xiang-ru

2017-07-01

Deep learning is a typical learning method widely studied in the fields of machine learning, pattern recognition, and artificial intelligence. This work investigates the problem of stellar atmospheric parameterization by constructing a deep neural network with five layers, and the node number in each layer of the network is respectively 3821-500-100-50-1. The proposed scheme is verified on both the real spectra measured by the Sloan Digital Sky Survey (SDSS) and the theoretic spectra computed with the Kurucz's New Opacity Distribution Function (NEWODF) model, to make an automatic estimation for three physical parameters: the effective temperature (Teff), surface gravitational acceleration (lg g), and metallic abundance (Fe/H). The results show that the stacked autoencoder deep neural network has a better accuracy for the estimation. On the SDSS spectra, the mean absolute errors (MAEs) are 79.95 for Teff/K, 0.0058 for (lg Teff/K), 0.1706 for lg (g/(cm·s-2)), and 0.1294 dex for the [Fe/H], respectively; On the theoretic spectra, the MAEs are 15.34 for Teff/K, 0.0011 for lg (Teff/K), 0.0214 for lg(g/(cm · s-2)), and 0.0121 dex for [Fe/H], respectively.

Recognition of emotions using multimodal physiological signals and an ensemble deep learning model.

Science.gov (United States)

Yin, Zhong; Zhao, Mengyuan; Wang, Yongxiong; Yang, Jingdong; Zhang, Jianhua

2017-03-01

Using deep-learning methodologies to analyze multimodal physiological signals becomes increasingly attractive for recognizing human emotions. However, the conventional deep emotion classifiers may suffer from the drawback of the lack of the expertise for determining model structure and the oversimplification of combining multimodal feature abstractions. In this study, a multiple-fusion-layer based ensemble classifier of stacked autoencoder (MESAE) is proposed for recognizing emotions, in which the deep structure is identified based on a physiological-data-driven approach. Each SAE consists of three hidden layers to filter the unwanted noise in the physiological features and derives the stable feature representations. An additional deep model is used to achieve the SAE ensembles. The physiological features are split into several subsets according to different feature extraction approaches with each subset separately encoded by a SAE. The derived SAE abstractions are combined according to the physiological modality to create six sets of encodings, which are then fed to a three-layer, adjacent-graph-based network for feature fusion. The fused features are used to recognize binary arousal or valence states. DEAP multimodal database was employed to validate the performance of the MESAE. By comparing with the best existing emotion classifier, the mean of classification rate and F-score improves by 5.26%. The superiority of the MESAE against the state-of-the-art shallow and deep emotion classifiers has been demonstrated under different sizes of the available physiological instances. Copyright © 2016 Elsevier Ireland Ltd. All rights reserved.

Applied model for the growth of the daytime mixed layer

DEFF Research Database (Denmark)

Batchvarova, E.; Gryning, Sven-Erik

1991-01-01

numerically. When the mixed layer is shallow or the atmosphere nearly neutrally stratified, the growth is controlled mainly by mechanical turbulence. When the layer is deep, its growth is controlled mainly by convective turbulence. The model is applied on a data set of the evolution of the height of the mixed...... layer in the morning hours, when both mechanical and convective turbulence contribute to the growth process. Realistic mixed-layer developments are obtained....

Development and application of deep convolutional neural network in target detection

Science.gov (United States)

Jiang, Xiaowei; Wang, Chunping; Fu, Qiang

2018-04-01

With the development of big data and algorithms, deep convolution neural networks with more hidden layers have more powerful feature learning and feature expression ability than traditional machine learning methods, making artificial intelligence surpass human level in many fields. This paper first reviews the development and application of deep convolutional neural networks in the field of object detection in recent years, then briefly summarizes and ponders some existing problems in the current research, and the future development of deep convolutional neural network is prospected.

Comparison of single and dual layer detector blocks for pre-clinical MRI–PET

International Nuclear Information System (INIS)

Thompson, Christopher; Stortz, Greg; Goertzen, Andrew; Berg, Eric; Retière, Fabrice; Kozlowski, Piotr; Ryner, Lawrence; Sossi, Vesna; Zhang, Xuezhu

2013-01-01

Dual or multi-layer crystal blocks have been proposed to minimise the radial blurring effect in PET scanners with small ring diameters. We measured two potential PET detector blocks' performance in a configuration which would allow 16 blocks in a ring which could be inserted in a small animal 7T MRI scanner. Two crystal sizes, 1.60×1.60 mm 2 and 1.20×1.20 mm 2 , were investigated. Single layer blocks had 10 or 12 mm deep crystals, the dual layer blocks had 4 mm deep crystals on the top layer and 6 mm deep crystals on the bottom layer. The crystals in the dual layer blocks are offset by ½ of the crystal pitch to allow for purely geometric crystal identification. Both were read out with SensL 4×4 SiPM arrays. The software identifies 64 crystals in the single layer and either 85 or 113 crystals in the dual layer array, (either 49 or 64 in the lower layers and 36 or 49 in the upper layers). All the crystals were clearly visible in the crystal identification images and their resolvability indexes (average FWHM/crystal separation) were shown to range from 0.29 for the best single layer block to 0.33 for the densest dual layer block. The best coincidence response FWHM was 0.95 mm for the densest block at the centre of the field. This degraded to 1.83 mm at a simulated radial offset of 16 mm from the centre, while the single layer crystals blurred this result to 3.4 mm. The energy resolution was 16.4±2.2% averaged over the 113 crystals of the densest block

Deep Learning for Detection of Object-Based Forgery in Advanced Video

Directory of Open Access Journals (Sweden)

Ye Yao

2017-12-01

Full Text Available Passive video forensics has drawn much attention in recent years. However, research on detection of object-based forgery, especially for forged video encoded with advanced codec frameworks, is still a great challenge. In this paper, we propose a deep learning-based approach to detect object-based forgery in the advanced video. The presented deep learning approach utilizes a convolutional neural network (CNN to automatically extract high-dimension features from the input image patches. Different from the traditional CNN models used in computer vision domain, we let video frames go through three preprocessing layers before being fed into our CNN model. They include a frame absolute difference layer to cut down temporal redundancy between video frames, a max pooling layer to reduce computational complexity of image convolution, and a high-pass filter layer to enhance the residual signal left by video forgery. In addition, an asymmetric data augmentation strategy has been established to get a similar number of positive and negative image patches before the training. The experiments have demonstrated that the proposed CNN-based model with the preprocessing layers has achieved excellent results.

Deep level defects in Ge-doped (010) β-Ga2O3 layers grown by plasma-assisted molecular beam epitaxy

Science.gov (United States)

Farzana, Esmat; Ahmadi, Elaheh; Speck, James S.; Arehart, Aaron R.; Ringel, Steven A.

2018-04-01

Deep level defects were characterized in Ge-doped (010) β-Ga2O3 layers grown by plasma-assisted molecular beam epitaxy (PAMBE) using deep level optical spectroscopy (DLOS) and deep level transient (thermal) spectroscopy (DLTS) applied to Ni/β-Ga2O3:Ge (010) Schottky diodes that displayed Schottky barrier heights of 1.50 eV. DLOS revealed states at EC - 2.00 eV, EC - 3.25 eV, and EC - 4.37 eV with concentrations on the order of 1016 cm-3, and a lower concentration level at EC - 1.27 eV. In contrast to these states within the middle and lower parts of the bandgap probed by DLOS, DLTS measurements revealed much lower concentrations of states within the upper bandgap region at EC - 0.1 - 0.2 eV and EC - 0.98 eV. There was no evidence of the commonly observed trap state at ˜EC - 0.82 eV that has been reported to dominate the DLTS spectrum in substrate materials synthesized by melt-based growth methods such as edge defined film fed growth (EFG) and Czochralski methods [Zhang et al., Appl. Phys. Lett. 108, 052105 (2016) and Irmscher et al., J. Appl. Phys. 110, 063720 (2011)]. This strong sensitivity of defect incorporation on crystal growth method and conditions is unsurprising, which for PAMBE-grown β-Ga2O3:Ge manifests as a relatively "clean" upper part of the bandgap. However, the states at ˜EC - 0.98 eV, EC - 2.00 eV, and EC - 4.37 eV are reminiscent of similar findings from these earlier results on EFG-grown materials, suggesting that possible common sources might also be present irrespective of growth method.

Importance of the variability of hydrographic preconditioning for deep convection in the Gulf of Lion, NW Mediterranean

Directory of Open Access Journals (Sweden)

L. Grignon

2010-06-01

Full Text Available We study the variability of hydrographic preconditioning defined as the heat and salt contents in the Ligurian Sea before convection. The stratification is found to reach a maximum in the intermediate layer in December, whose causes and consequences for the interannual variability of convection are investigated. Further study of the interannual variability and correlation tests between the properties of the deep water formed and the winter surface fluxes support the description of convection as a process that transfers the heat and salt contents from the top and intermediate layers to the deep layer. A proxy for the rate of transfer is given by the final convective mixed layer depth, that is shown to depend equally on the surface fluxes and on the preconditioning. In particular, it is found that deep convection in winter 2004–2005 would have happened even with normal winter conditions, due to low pre-winter stratification.

Deep UV Native Fluorescence Imaging of Antarctic Cryptoendolithic Communities

Science.gov (United States)

Storrie-Lombardi, M. C.; Douglas, S.; Sun, H.; McDonald, G. D.; Bhartia, R.; Nealson, K. H.; Hug, W. F.

2001-01-01

An interdisciplinary team at the Jet Propulsion Laboratory Center for Life Detection has embarked on a project to provide in situ chemical and morphological characterization of Antarctic cryptoendolithic microbial communities. We present here in situ deep ultraviolet (UV) native fluorescence and environmental scanning electron microscopy images transiting 8.5 mm into a sandstone sample from the Antarctic Dry Valleys. The deep ultraviolet imaging system employs 224.3, 248.6, and 325 nm lasers to elicit differential fluorescence and resonance Raman responses from biomolecules and minerals. The 224.3 and 248.6 nm lasers elicit a fluorescence response from the aromatic amino and nucleic acids. Excitation at 325 nm may elicit activity from a variety of biomolecules, but is more likely to elicit mineral fluorescence. The resultant fluorescence images provide in situ chemical and morphological maps of microorganisms and the associated organic matrix. Visible broadband reflectance images provide orientation against the mineral background. Environmental scanning electron micrographs provided detailed morphological information. The technique has made possible the construction of detailed fluorescent maps extending from the surface of an Antarctic sandstone sample to a depth of 8.5 mm. The images detect no evidence of microbial life in the superficial 0.2 mm crustal layer. The black lichen component between 0.3 and 0.5 mm deep absorbs all wavelengths of both laser and broadband illumination. Filamentous deep ultraviolet native fluorescent activity dominates in the white layer between 0.6 mm and 5.0 mm from the surface. These filamentous forms are fungi that continue into the red (iron-rich) region of the sample extending from 5.0 to 8.5 mm. Using differential image subtraction techniques it is possible to identify fungal nuclei. The ultraviolet response is markedly attenuated in this region, apparently from the absorption of ultraviolet light by iron-rich particles coating

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs.

Science.gov (United States)

Chen, Liang-Chieh; Papandreou, George; Kokkinos, Iasonas; Murphy, Kevin; Yuille, Alan L

2018-04-01

In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First, we highlight convolution with upsampled filters, or 'atrous convolution', as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second, we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third, we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed "DeepLab" system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7 percent mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.

Soft-Deep Boltzmann Machines

OpenAIRE

Kiwaki, Taichi

2015-01-01

We present a layered Boltzmann machine (BM) that can better exploit the advantages of a distributed representation. It is widely believed that deep BMs (DBMs) have far greater representational power than its shallow counterpart, restricted Boltzmann machines (RBMs). However, this expectation on the supremacy of DBMs over RBMs has not ever been validated in a theoretical fashion. In this paper, we provide both theoretical and empirical evidences that the representational power of DBMs can be a...

Plant Species Identification by Bi-channel Deep Convolutional Networks

Science.gov (United States)

He, Guiqing; Xia, Zhaoqiang; Zhang, Qiqi; Zhang, Haixi; Fan, Jianping

2018-04-01

Plant species identification achieves much attention recently as it has potential application in the environmental protection and human life. Although deep learning techniques can be directly applied for plant species identification, it still needs to be designed for this specific task to obtain the state-of-art performance. In this paper, a bi-channel deep learning framework is developed for identifying plant species. In the framework, two different sub-networks are fine-tuned over their pretrained models respectively. And then a stacking layer is used to fuse the output of two different sub-networks. We construct a plant dataset of Orchidaceae family for algorithm evaluation. Our experimental results have demonstrated that our bi-channel deep network can achieve very competitive performance on accuracy rates compared to the existing deep learning algorithm.

Adapted ECC ozonesonde for long-duration flights aboard boundary-layer pressurised balloons

Science.gov (United States)

Gheusi, François; Durand, Pierre; Verdier, Nicolas; Dulac, François; Attié, Jean-Luc; Commun, Philippe; Barret, Brice; Basdevant, Claude; Clenet, Antoine; Derrien, Solène; Doerenbecher, Alexis; El Amraoui, Laaziz; Fontaine, Alain; Hache, Emeric; Jambert, Corinne; Jaumouillé, Elodie; Meyerfeld, Yves; Roblou, Laurent; Tocquer, Flore

2016-12-01

Since the 1970s, the French space agency CNES has developed boundary-layer pressurised balloons (BLPBs) with the capability to transport lightweight scientific payloads at isopycnic level and offer a quasi-Lagrangian sampling of the lower atmosphere over very long distances and durations (up to several weeks).Electrochemical concentration cell (ECC) ozonesondes are widely used under small sounding balloons. However, their autonomy is limited to a few hours owing to power consumption and electrolyte evaporation. An adaptation of the ECC sonde has been developed specifically for long-duration BLPB flights. Compared to conventional ECC sondes, the main feature is the possibility of programming periodic measurement sequences (with possible remote control during the flight). To increase the ozonesonde autonomy, the strategy has been adopted of short measurement sequences (2-3 min) regularly spaced in time (e.g. every 15 min). The rest of the time, the sonde pump is turned off. Results of preliminary ground-based tests are first presented. In particular, the sonde was able to provide correct ozone concentrations against a reference UV-absorption ozone analyser every 15 min for 4 days. Then we illustrate results from 16 BLBP flights launched over the western Mediterranean during three summer field campaigns of the ChArMEx project (http://charmex.lsce.ipsl.fr): TRAQA in 2012, and ADRIMED and SAFMED in 2013. BLPB drifting altitudes were in the range 0.25-3.2 km. The longest flight lasted more than 32 h and covered more than 1000 km. Satisfactory data were obtained when compared to independent ozone measurements close in space and time. The quasi-Lagrangian measurements allowed a first look at ozone diurnal evolution in the marine boundary layer as well as in the lower free troposphere. During some flight segments, there was indication of photochemical ozone production in the marine boundary layer or even in the free troposphere, at rates ranging from 1 to 2 ppbv h -1, which

A deep-level transient spectroscopy study of gamma-ray irradiation on the passivation properties of silicon nitride layer on silicon

Science.gov (United States)

Dong, Peng; Yu, Xuegong; Ma, Yao; Xie, Meng; Li, Yun; Huang, Chunlai; Li, Mo; Dai, Gang; Zhang, Jian

2017-08-01

Plasma-enhanced chemical vapor deposited silicon nitride (SiNx) films are extensively used as passivation material in the solar cell industry. Such SiNx passivation layers are the most sensitive part to gamma-ray irradiation in solar cells. In this work, deep-level transient spectroscopy has been applied to analyse the influence of gamma-ray irradiation on the passivation properties of SiNx layer on silicon. It is shown that the effective carrier lifetime decreases with the irradiation dose. At the same time, the interface state density is significantly increased after irradiation, and its energy distribution is broadened and shifts deeper with respect to the conduction band edge, which makes the interface states becoming more efficient recombination centers for carriers. Besides, C-V characteristics show a progressive negative shift with increasing dose, indicating the generation of effective positive charges in SiNx films. Such positive charges are beneficial for shielding holes from the n-type silicon substrates, i. e. the field-effect passivation. However, based on the reduced carrier lifetime after irradiation, it can be inferred that the irradiation induced interface defects play a dominant role over the trapped positive charges, and therefore lead to the degradation of passivation properties of SiNx on silicon.

Deep learning—Accelerating Next Generation Performance Analysis Systems?

Directory of Open Access Journals (Sweden)

Heike Brock

2018-02-01

Full Text Available Deep neural network architectures show superior performance in recognition and prediction tasks of the image, speech and natural language domains. The success of such multi-layered networks encourages their implementation in further application scenarios as the retrieval of relevant motion information for performance enhancement in sports. However, to date deep learning is only seldom applied to activity recognition problems of the human motion domain. Therefore, its use for sports data analysis might remain abstract to many practitioners. This paper provides a survey on recent works in the field of high-performance motion data and examines relevant technologies for subsequent deployment in real training systems. In particular, it discusses aspects of data acquisition, processing and network modeling. Analysis suggests the advantage of deep neural networks under difficult and noisy data conditions. However, further research is necessary to confirm the benefit of deep learning for next generation performance analysis systems.

Deep Learning for Plant Identification in Natural Environment.

Science.gov (United States)

Sun, Yu; Liu, Yuan; Wang, Guan; Zhang, Haiyan

2017-01-01

Plant image identification has become an interdisciplinary focus in both botanical taxonomy and computer vision. The first plant image dataset collected by mobile phone in natural scene is presented, which contains 10,000 images of 100 ornamental plant species in Beijing Forestry University campus. A 26-layer deep learning model consisting of 8 residual building blocks is designed for large-scale plant classification in natural environment. The proposed model achieves a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.

Deep Learning for Plant Identification in Natural Environment

Directory of Open Access Journals (Sweden)

Yu Sun

2017-01-01

Full Text Available Plant image identification has become an interdisciplinary focus in both botanical taxonomy and computer vision. The first plant image dataset collected by mobile phone in natural scene is presented, which contains 10,000 images of 100 ornamental plant species in Beijing Forestry University campus. A 26-layer deep learning model consisting of 8 residual building blocks is designed for large-scale plant classification in natural environment. The proposed model achieves a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.

Double seismic zone for deep earthquakes in the izu-bonin subduction zone.

Science.gov (United States)

Iidaka, T; Furukawa, Y

1994-02-25

A double seismic zone for deep earthquakes was found in the Izu-Bonin region. An analysis of SP-converted phases confirms that the deep seismic zone consists of two layers separated by approximately 20 kilometers. Numerical modeling of the thermal structure implies that the hypocenters are located along isotherms of 500 degrees to 550 degrees C, which is consistent with the hypothesis that deep earthquakes result from the phase transition of metastable olivine to a high-pressure phase in the subducting slab.

Global assessment of benthic nepheloid layers and linkage with upper ocean dynamics

Science.gov (United States)

Gardner, Wilford D.; Richardson, Mary Jo; Mishonov, Alexey V.

2018-01-01

Global maps of the maximum bottom concentration, thickness, and integrated particle mass in benthic nepheloid layers are published here to support collaborations to understand deep ocean sediment dynamics, linkage with upper ocean dynamics, and assessing the potential for scavenging of adsorption-prone elements near the deep ocean seafloor. Mapping the intensity of benthic particle concentrations from natural oceanic processes also provides a baseline that will aid in quantifying the industrial impact of current and future deep-sea mining. Benthic nepheloid layers have been mapped using 6,392 full-depth profiles made during 64 cruises using our transmissometers mounted on CTDs in multiple national/international programs including WOCE, SAVE, JGOFS, CLIVAR-Repeat Hydrography, and GO-SHIP during the last four decades. Intense benthic nepheloid layers are found in areas where eddy kinetic energy in overlying waters, mean kinetic energy 50 m above bottom (mab), and energy dissipation in the bottom boundary layer are near the highest values in the ocean. Areas of intense benthic nepheloid layers include the Western North Atlantic, Argentine Basin in the South Atlantic, parts of the Southern Ocean and areas around South Africa. Benthic nepheloid layers are weak or absent in most of the Pacific, Indian, and Atlantic basins away from continental margins. High surface eddy kinetic energy is associated with the Kuroshio Current east of Japan. Data south of the Kuroshio show weak nepheloid layers, but no transmissometer data exist beneath the Kuroshio, a deficiency that should be remedied to increase understanding of eddy dynamics in un-sampled and under-sampled oceanic areas.

CFD Simulation of Flow Tones from Grazing Flow past a Deep Cavity

International Nuclear Information System (INIS)

T Bagwell

2006-01-01

Locked-in flow tones due to shear flow over a deep cavity are investigated using Large Eddy Simulation (LES). An isentropic form of the compressible Navier-Stokes equations (pseudo-compressibility) is used to couple the vertical flow over the cavity mouth with the deep cavity resonances (1). Comparisons to published experimental data (2) show that the pseudo-compressible LES formulation is capable of predicting the feedforward excitation of the deep cavity resonator, as well as the feedback process from the resonator to the flow source. By systematically increasing the resonator damping level, it is shown that strong lock-in results in a more organized shear layer than is observed for the locked-out flow state. By comparison, weak interactions (non-locked-in) produce no change in the shear layer characteristics. This supports the 40 dB definition of lock-in defined in the experiment

Persistent photoconductivity in AlGaN/GaN heterojunction channels caused by the ionization of deep levels in the AlGaN barrier layer

International Nuclear Information System (INIS)

Murayama, H.; Akiyama, Y.; Niwa, R.; Sakashita, H.; Sakaki, H.; Kachi, T.; Sugimoto, M.

2013-01-01

Time-dependent responses of drain current (I d ) in an AlGaN/GaN HEMT under UV (3.3 eV) and red (2.0 eV) light illumination have been studied at 300 K and 250 K. UV illumination enhances I d by about 10 %, indicating that the density of two-dimensional electrons is raised by about 10 12 cm −2 . When UV light is turned off at 300 K, a part of increased I d decays quickly but the other part of increment is persistent, showing a slow decay. At 250 K, the majority of increment remains persistent. It is found that such a persistent increase of I d at 250 K can be partially erased by the illumination of red light. These photo-responses are explained by a simple band-bending model in which deep levels in the AlGaN barrier get positively charged by the UV light, resulting in a parabolic band bending in the AlGaN layer, while some potion of those deep levels are neutralized by the red light

Multiscale deep features learning for land-use scene recognition

Science.gov (United States)

Yuan, Baohua; Li, Shijin; Li, Ning

2018-01-01

The features extracted from deep convolutional neural networks (CNNs) have shown their promise as generic descriptors for land-use scene recognition. However, most of the work directly adopts the deep features for the classification of remote sensing images, and does not encode the deep features for improving their discriminative power, which can affect the performance of deep feature representations. To address this issue, we propose an effective framework, LASC-CNN, obtained by locality-constrained affine subspace coding (LASC) pooling of a CNN filter bank. LASC-CNN obtains more discriminative deep features than directly extracted from CNNs. Furthermore, LASC-CNN builds on the top convolutional layers of CNNs, which can incorporate multiscale information and regions of arbitrary resolution and sizes. Our experiments have been conducted using two widely used remote sensing image databases, and the results show that the proposed method significantly improves the performance when compared to other state-of-the-art methods.

Surface Improvement of Shafts by Turn-Assisted Deep Cold Rolling Process

Directory of Open Access Journals (Sweden)

Prabhu Raghavendra

2016-01-01

Full Text Available It is well recognized that mechanical surface enhancement methods can significantly improve the characteristics of highly-stressed metallic components. Deep cold rolling is one of such technique which is particularly attractive since it is possible to generate, near the surface, deep compressive residual stresses and work hardened layers while retaining a relatively smooth surface finish. In this paper, the effect of turn-assisted deep cold rolling on AISI 4140 steel is examined, with emphasis on the residual stress state. Based on the X-ray diffraction measurements, it is found that turn-assisted deep cold rolling can be quite effective in retarding the initiation and initial propagation of fatigue cracks in AISI 4140 steel.

A data-driven multi-model methodology with deep feature selection for short-term wind forecasting

International Nuclear Information System (INIS)

Feng, Cong; Cui, Mingjian; Hodge, Bri-Mathias; Zhang, Jie

2017-01-01

Highlights: • An ensemble model is developed to produce both deterministic and probabilistic wind forecasts. • A deep feature selection framework is developed to optimally determine the inputs to the forecasting methodology. • The developed ensemble methodology has improved the forecasting accuracy by up to 30%. - Abstract: With the growing wind penetration into the power system worldwide, improving wind power forecasting accuracy is becoming increasingly important to ensure continued economic and reliable power system operations. In this paper, a data-driven multi-model wind forecasting methodology is developed with a two-layer ensemble machine learning technique. The first layer is composed of multiple machine learning models that generate individual forecasts. A deep feature selection framework is developed to determine the most suitable inputs to the first layer machine learning models. Then, a blending algorithm is applied in the second layer to create an ensemble of the forecasts produced by first layer models and generate both deterministic and probabilistic forecasts. This two-layer model seeks to utilize the statistically different characteristics of each machine learning algorithm. A number of machine learning algorithms are selected and compared in both layers. This developed multi-model wind forecasting methodology is compared to several benchmarks. The effectiveness of the proposed methodology is evaluated to provide 1-hour-ahead wind speed forecasting at seven locations of the Surface Radiation network. Numerical results show that comparing to the single-algorithm models, the developed multi-model framework with deep feature selection procedure has improved the forecasting accuracy by up to 30%.

Irrigation and Nitrogen Regimes Promote the Use of Soil Water and Nitrate Nitrogen from Deep Soil Layers by Regulating Root Growth in Wheat.

Science.gov (United States)

Liu, Weixing; Ma, Geng; Wang, Chenyang; Wang, Jiarui; Lu, Hongfang; Li, Shasha; Feng, Wei; Xie, Yingxin; Ma, Dongyun; Kang, Guozhang

2018-01-01

Unreasonably high irrigation levels and excessive nitrogen (N) supplementation are common occurrences in the North China Plain that affect winter wheat production. Therefore, a 6-yr-long stationary field experiment was conducted to investigate the effects of irrigation and N regimes on root development and their relationship with soil water and N use in different soil layers. Compared to the non-irrigated treatment (W0), a single irrigation at jointing (W1) significantly increased yield by 3.6-45.6%. With increases in water (W2, a second irrigation at flowering), grain yield was significantly improved by 14.1-45.3% compared to the W1 treatments during the drier growing seasons (2010-2011, 2012-2013, and 2015-2016). However, under sufficient pre-sowing soil moisture conditions, grain yield was not increased, and water use efficiency (WUE) decreased significantly in the W2 treatments during normal precipitation seasons (2011-2012, 2013-2014, and 2014-2015). Irrigating the soil twice inhibited root growth into the deeper soil depth profiles and thus weakened the utilization of soil water and NO 3 -N from the deep soil layers. N applications increased yield by 19.1-64.5%, with a corresponding increase in WUE of 66.9-83.9% compared to the no-N treatment (N0). However, there was no further increase in grain yield and the WUE response when N rates exceeded 240 and 180 kg N ha -1 , respectively. A N application rate of 240 kg ha -1 facilitated root growth in the deep soil layers, which was conducive to utilization of soil water and NO 3 -N and also in reducing the residual NO 3 -N. Correlation analysis indicated that the grain yield was significantly positively correlated with soil water storage (SWS) and nitrate nitrogen accumulation (SNA) prior to sowing. Therefore, N rates of 180-240 kg ha -1 with two irrigations can reduce the risk of yield loss that occurs due to reduced precipitation during the wheat growing seasons, while under better soil moisture conditions, a

Quantum state engineering with ultra-short-period (AlN)m/(GaN)n superlattices for narrowband deep-ultraviolet detection.

Science.gov (United States)

Gao, Na; Lin, Wei; Chen, Xue; Huang, Kai; Li, Shuping; Li, Jinchai; Chen, Hangyang; Yang, Xu; Ji, Li; Yu, Edward T; Kang, Junyong

2014-12-21

Ultra-short-period (AlN)m/(GaN)n superlattices with tunable well and barrier atomic layer numbers were grown by metal-organic vapour phase epitaxy, and employed to demonstrate narrowband deep ultraviolet photodetection. High-resolution transmission electron microscopy and X-ray reciprocal space mapping confirm that superlattices containing well-defined, coherently strained GaN and AlN layers as thin as two atomic layers (∼ 0.5 nm) were grown. Theoretical and experimental results demonstrate that an optical absorption band as narrow as 9 nm (210 meV) at deep-ultraviolet wavelengths can be produced, and is attributable to interband transitions between quantum states along the [0001] direction in ultrathin GaN atomic layers isolated by AlN barriers. The absorption wavelength can be precisely engineered by adjusting the thickness of the GaN atomic layers because of the quantum confinement effect. These results represent a major advance towards the realization of wavelength selectable and narrowband photodetectors in the deep-ultraviolet region without any additional optical filters.

Interactions between deep bedrock aquifers and surface water in function of recharge and topography: a numerical study

Science.gov (United States)

Goderniaux, P.; Davy, P.; Le Borgne, T.; Bresciani, E.; Jimenez-Martinez, J.

2011-12-01

In crystalline rock regions, such as Brittany (France), important reserves of groundwater into deep fractured aquifers are increasingly used and provide high quality water compared to shallow aquifers which can be subject to agricultural contamination. However, recharge processes of these deep aquifers and interactions with surface water are not yet fully understood. In some areas, intensive pumping is carried out without guarantee of the resource quantity and quality. Understanding these processes is crucial for sustainable management of the resource. In this study, we study how deep groundwater fluxes, pathways, ages, and river-aquifer interactions vary according to recharge. We assume that water flowing from the ground surface is distributed between shallow more permeable layers and deep layers. This repartition mostly depends on recharge rates. With high recharge, groundwater levels are high and subsurface streamlines are relatively short between recharge areas and existing draining rivers, which constitutes a very dense network. Therefore, most of the groundwater fluxes occur through the more permeable shallow layers. With low recharge, groundwater levels are lower, and river and shallow permeable levels are partly disconnected from each other. This induces a general increase of the groundwater streamlines length from the recharge areas to more sporadic discharge areas, and more fluxes occur through the deep layers. Recharge conditions and river-aquifer interactions have changed over the last thousands of years, due to change in precipitation, temperatures, existence of permafrost, etc. They have strongly influenced deep groundwater fluxes and can explain current groundwater age and flux distribution. To study these interactions, a regional-scale finite difference flow model was implemented. The model covers an area of 1400 km 2 , a depth of 1 km, and the topography is characteristic of Brittany. As rivers are mainly fed by groundwater drainage, seepages faces

Deep-Learning-Based Approach for Prediction of Algal Blooms

Directory of Open Access Journals (Sweden)

Feng Zhang

2016-10-01

Full Text Available Algal blooms have recently become a critical global environmental concern which might put economic development and sustainability at risk. However, the accurate prediction of algal blooms remains a challenging scientific problem. In this study, a novel prediction approach for algal blooms based on deep learning is presented—a powerful tool to represent and predict highly dynamic and complex phenomena. The proposed approach constructs a five-layered model to extract detailed relationships between the density of phytoplankton cells and various environmental parameters. The algal blooms can be predicted by the phytoplankton density obtained from the output layer. A case study is conducted in coastal waters of East China using both our model and a traditional back-propagation neural network for comparison. The results show that the deep-learning-based model yields better generalization and greater accuracy in predicting algal blooms than a traditional shallow neural network does.

A Deep-Structured Conditional Random Field Model for Object Silhouette Tracking.

Directory of Open Access Journals (Sweden)

Mohammad Javad Shafiee

Full Text Available In this work, we introduce a deep-structured conditional random field (DS-CRF model for the purpose of state-based object silhouette tracking. The proposed DS-CRF model consists of a series of state layers, where each state layer spatially characterizes the object silhouette at a particular point in time. The interactions between adjacent state layers are established by inter-layer connectivity dynamically determined based on inter-frame optical flow. By incorporate both spatial and temporal context in a dynamic fashion within such a deep-structured probabilistic graphical model, the proposed DS-CRF model allows us to develop a framework that can accurately and efficiently track object silhouettes that can change greatly over time, as well as under different situations such as occlusion and multiple targets within the scene. Experiment results using video surveillance datasets containing different scenarios such as occlusion and multiple targets showed that the proposed DS-CRF approach provides strong object silhouette tracking performance when compared to baseline methods such as mean-shift tracking, as well as state-of-the-art methods such as context tracking and boosted particle filtering.

Flood frequency matters: Why climate change degrades deep-water quality of peri-alpine lakes

Science.gov (United States)

Fink, Gabriel; Wessels, Martin; Wüest, Alfred

2016-09-01

Sediment-laden riverine floods transport large quantities of dissolved oxygen into the receiving deep layers of lakes. Hence, the water quality of deep lakes is strongly influenced by the frequency of riverine floods. Although flood frequency reflects climate conditions, the effects of climate variability on the water quality of deep lakes is largely unknown. We quantified the effects of climate variability on the potential shifts in the flood regime of the Alpine Rhine, the main catchment of Lake Constance, and determined the intrusion depths of riverine density-driven underflows and the subsequent effects on water exchange rates in the lake. A simplified hydrodynamic underflow model was developed and validated with observed river inflow and underflow events. The model was implemented to estimate underflow statistics for different river inflow scenarios. Using this approach, we integrated present and possible future flood frequencies to underflow occurrences and intrusion depths in Lake Constance. The results indicate that more floods will increase the number of underflows and the intensity of deep-water renewal - and consequently will cause higher deep-water dissolved oxygen concentrations. Vice versa, fewer floods weaken deep-water renewal and lead to lower deep-water dissolved oxygen concentrations. Meanwhile, a change from glacial nival regime (present) to a nival pluvial regime (future) is expected to decrease deep-water renewal. While flood frequencies are not expected to change noticeably for the next decades, it is most likely that increased winter discharge and decreased summer discharge will reduce the number of deep density-driven underflows by 10% and favour shallower riverine interflows in the upper hypolimnion. The renewal in the deepest layers is expected to be reduced by nearly 27%. This study underlines potential consequences of climate change on the occurrence of deep river underflows and water residence times in deep lakes.

Factors governing the deep ventilation of the Red Sea

KAUST Repository

Papadopoulos, Vassilis P.

2015-11-19

A variety of data based on hydrographic measurements, satellite observations, reanalysis databases, and meteorological observations are used to explore the interannual variability and factors governing the deep water formation in the northern Red Sea. Historical and recent hydrographic data consistently indicate that the ventilation of the near-bottom layer in the Red Sea is a robust feature of the thermohaline circulation. Dense water capable to reach the bottom layers of the Red Sea can be regularly produced mostly inside the Gulfs of Aqaba and Suez. Occasionally, during colder than usual winters, deep water formation may also take place over coastal areas in the northernmost end of the open Red Sea just outside the Gulfs of Aqaba and Suez. However, the origin as well as the amount of deep waters exhibit considerable interannual variability depending not only on atmospheric forcing but also on the water circulation over the northern Red Sea. Analysis of several recent winters shows that the strength of the cyclonic gyre prevailing in the northernmost part of the basin can effectively influence the sea surface temperature (SST) and intensify or moderate the winter surface cooling. Upwelling associated with periods of persistent gyre circulation lowers the SST over the northernmost part of the Red Sea and can produce colder than normal winter SST even without extreme heat loss by the sea surface. In addition, the occasional persistence of the cyclonic gyre feeds the surface layers of the northern Red Sea with nutrients, considerably increasing the phytoplankton biomass.

Bidirectional Nonnegative Deep Model and Its Optimization in Learning

Directory of Open Access Journals (Sweden)

Xianhua Zeng

2016-01-01

Full Text Available Nonnegative matrix factorization (NMF has been successfully applied in signal processing as a simple two-layer nonnegative neural network. Projective NMF (PNMF with fewer parameters was proposed, which projects a high-dimensional nonnegative data onto a lower-dimensional nonnegative subspace. Although PNMF overcomes the problem of out-of-sample of NMF, it does not consider the nonlinear characteristic of data and is only a kind of narrow signal decomposition method. In this paper, we combine the PNMF with deep learning and nonlinear fitting to propose a bidirectional nonnegative deep learning (BNDL model and its optimization learning algorithm, which can obtain nonlinear multilayer deep nonnegative feature representation. Experiments show that the proposed model can not only solve the problem of out-of-sample of NMF but also learn hierarchical nonnegative feature representations with better clustering performance than classical NMF, PNMF, and Deep Semi-NMF algorithms.

Controls on deep drainage beneath the root soil zone in snowmelt-dominated environments

Science.gov (United States)

Hammond, J. C.; Harpold, A. A.; Kampf, S. K.

2017-12-01

Snowmelt is the dominant source of streamflow generation and groundwater recharge in many high elevation and high latitude locations, yet we still lack a detailed understanding of how snowmelt is partitioned between the soil, deep drainage, and streamflow under a variety of soil, climate, and snow conditions. Here we use Hydrus 1-D simulations with historical inputs from five SNOTEL snow monitoring sites in each of three regions, Cascades, Sierra, and Southern Rockies, to investigate how inter-annual variability on water input rate and duration affects soil saturation and deep drainage. Each input scenario was run with three different soil profiles of varying hydraulic conductivity, soil texture, and bulk density. We also created artificial snowmelt scenarios to test how snowmelt intermittence affects deep drainage. Results indicate that precipitation is the strongest predictor (R2 = 0.83) of deep drainage below the root zone, with weaker relationships observed between deep drainage and snow persistence, peak snow water equivalent, and melt rate. The ratio of deep drainage to precipitation shows a stronger positive relationship to melt rate suggesting that a greater fraction of input becomes deep drainage at higher melt rates. For a given amount of precipitation, rapid, concentrated snowmelt may create greater deep drainage below the root zone than slower, intermittent melt. Deep drainage requires saturation below the root zone, so saturated hydraulic conductivity serves as a primary control on deep drainage magnitude. Deep drainage response to climate is mostly independent of soil texture because of its reliance on saturated conditions. Mean water year saturations of deep soil layers can predict deep drainage and may be a useful way to compare sites in soils with soil hydraulic porosities. The unit depth of surface runoff often is often greater than deep drainage at daily and annual timescales, as snowmelt exceeds infiltration capacity in near-surface soil layers

Modeling Language and Cognition with Deep Unsupervised Learning:A Tutorial Overview

OpenAIRE

Marco eZorzi; Marco eZorzi; Alberto eTestolin; Ivilin Peev Stoianov; Ivilin Peev Stoianov

2013-01-01

Deep unsupervised learning in stochastic recurrent neural networks with many layers of hidden units is a recent breakthrough in neural computation research. These networks build a hierarchy of progressively more complex distributed representations of the sensory data by fitting a hierarchical generative model. In this article we discuss the theoretical foundations of this approach and we review key issues related to training, testing and analysis of deep networks for modeling language and cog...

Modeling language and cognition with deep unsupervised learning: a tutorial overview

OpenAIRE

Zorzi, Marco; Testolin, Alberto; Stoianov, Ivilin P.

2013-01-01

Deep unsupervised learning in stochastic recurrent neural networks with many layers of hidden units is a recent breakthrough in neural computation research. These networks build a hierarchy of progressively more complex distributed representations of the sensory data by fitting a hierarchical generative model. In this article we discuss the theoretical foundations of this approach and we review key issues related to training, testing and analysis of deep networks for modeling language and cog...

Relative Humidity in the Tropopause Saturation Layer

Science.gov (United States)

Selkirk, H. B.; Schoeberl, M. R.; Pfister, L.; Thornberry, T. D.; Bui, T. V.

2017-12-01

The tropical tropopause separates two very different atmospheric regimes: the stable lower stratosphere where the air is both extremely dry and nearly always so, and a transition layer in the uppermost tropical troposphere, where humidity on average increases rapidly downward but can undergo substantial temporal fluctuations. The processes that control the humidity in this layer below the tropopause include convective detrainment (which can result in either a net hydration or dehydration), slow ascent, wave motions and advection. Together these determine the humidity of the air that eventually passes through the tropopause and into the stratosphere, and we refer to this layer as the tropopause saturation layer or TSL. We know from in situ water vapor observations such as Ticosonde's 12-year balloonsonde record at Costa Rica that layers of supersaturation are frequently observed in the TSL. While their frequency is greatest during the local rainy season from June through October, supersaturation is also observed in the boreal winter dry season when deep convection is well south of Costa Rica. In other words, local convection is not a necessary condition for the presence of supersaturation. Furthermore, there are indications from airborne measurements during the recent POSIDON campaign at Guam that if anything deep convection tends to `reset' the TSL locally to a state of just-saturation. Conversely, it may be that layers of supersaturation are the result of slow ascent. To explore these ideas we take Ticosonde water vapor observations from the TSL, stratify them on the basis of relative humidity and report on the differences in the the history of upstream convective influence between supersaturated parcels and those that are not.

Effect of inversion layer at iron pyrite surface on photovoltaic device

Science.gov (United States)

Uchiyama, Shunsuke; Ishikawa, Yasuaki; Uraoka, Yukiharu

2018-03-01

Iron pyrite has great potential as a thin-film solar cell material because it has high optical absorption, low cost, and is earth-abundant. However, previously reported iron pyrite solar cells showed poor photovoltaic characteristics. Here, we have numerically simulated its photovoltaic characteristics and band structures by utilizing a two-dimensional (2D) device simulator, ATLAS, to evaluate the effects of an inversion layer at the surface and a high density of deep donor defect states in the bulk. We found that previous device structures did not consider the inversion layer at the surface region of iron pyrite, which made it difficult to obtain the conversion efficiency. Therefore, we remodeled the device structure and suggested that removing the inversion layer and reducing the density of deep donor defect states would lead to a high conversion efficiency of iron pyrite solar cells.

Realization of Chinese word segmentation based on deep learning method

Science.gov (United States)

Wang, Xuefei; Wang, Mingjiang; Zhang, Qiquan

2017-08-01

In recent years, with the rapid development of deep learning, it has been widely used in the field of natural language processing. In this paper, I use the method of deep learning to achieve Chinese word segmentation, with large-scale corpus, eliminating the need to construct additional manual characteristics. In the process of Chinese word segmentation, the first step is to deal with the corpus, use word2vec to get word embedding of the corpus, each character is 50. After the word is embedded, the word embedding feature is fed to the bidirectional LSTM, add a linear layer to the hidden layer of the output, and then add a CRF to get the model implemented in this paper. Experimental results show that the method used in the 2014 People's Daily corpus to achieve a satisfactory accuracy.

Dro1, a major QTL involved in deep rooting of rice under upland field conditions.

Science.gov (United States)

Uga, Yusaku; Okuno, Kazutoshi; Yano, Masahiro

2011-05-01

Developing a deep root system is an important strategy for avoiding drought stress in rice. Using the 'basket' method, the ratio of deep rooting (RDR; the proportion of total roots that elongated through the basket bottom) was calculated to evaluate deep rooting. A new major quantitative trait locus (QTL) controlling RDR was detected on chromosome 9 by using 117 recombinant inbred lines (RILs) derived from a cross between the lowland cultivar IR64, with shallow rooting, and the upland cultivar Kinandang Patong (KP), with deep rooting. This QTL explained 66.6% of the total phenotypic variance in RDR in the RILs. A BC(2)F(3) line homozygous for the KP allele of the QTL had an RDR of 40.4%, compared with 2.6% for the homozygous IR64 allele. Fine mapping of this QTL was undertaken using eight BC(2)F(3) recombinant lines. The RDR QTL Dro1 (Deeper rooting 1) was mapped between the markers RM24393 and RM7424, which delimit a 608.4 kb interval in the reference cultivar Nipponbare. To clarify the influence of Dro1 in an upland field, the root distribution in different soil layers was quantified by means of core sampling. A line homozygous for the KP allele of Dro1 (Dro1-KP) and IR64 did not differ in root dry weight in the shallow soil layers (0-25 cm), but root dry weight of Dro1-KP in deep soil layers (25-50 cm) was significantly greater than that of IR64, suggesting that Dro1 plays a crucial role in increased deep rooting under upland field conditions.

Tephrostratigraphy the DEEP site record, Lake Ohrid

Science.gov (United States)

Leicher, N.; Zanchetta, G.; Sulpizio, R.; Giaccio, B.; Wagner, B.; Francke, A.

2016-12-01

In the central Mediterranean region, tephrostratigraphy has been proofed to be a suitable and powerful tool for dating and correlating marine and terrestrial records. However, for the period older 200 ka, tephrostratigraphy is incomplete and restricted to some Italian continental basins (e.g. Sulmona, Acerno, Mercure), and continuous records downwind of the Italian volcanoes are rare. Lake Ohrid (Macedonia/Albania) in the eastern Mediterranean region fits this requisite and is assumed to be the oldest continuously existing lake of Europe. A continous record (DEEP) was recovered within the scope of the ICDP deep-drilling campaign SCOPSCO (Scientific Collaboration on Past Speciation Conditions in Lake Ohrid). In the uppermost 450 meters of the record, covering more than 1.2 Myrs of Italian volcanism, 54 tephra layers were identified during core-opening and description. A first tephrostratigraphic record was established for the uppermost 248 m ( 637 ka). Major element analyses (EDS/WDS) were carried out on juvenile glass fragments and 15 out of 35 tephra layers have been identified and correlated with known and dated eruptions of Italian volcanoes. Existing 40Ar/39Ar ages were re-calculated by using the same flux standard and used as first order tie points to develop a robust chronology for the DEEP site succession. Between 248 and 450 m of the DEEP site record, another 19 tephra horizons were identified and are subject of ongoing work. These deposits, once correlated with known and dated tephra, will hopefully enable dating this part of the succession, likely supported by major paleomagnetic events, such as the Brunhes-Matuyama boundary, or the Cobb-Mountain or the Jaramillo excursions. This makes the Lake Ohrid record a unique continuous, distal record of Italian volcanic activity, which is candidate to become the template for the central Mediterranean tephrostratigraphy, especially for the hitherto poorly known and explored lower Middle Pleistocene period.

Late Eocene impact events recorded in deep-sea sediments

Science.gov (United States)

Glass, B. P.

1988-01-01

Raup and Sepkoski proposed that mass extinctions have occurred every 26 Myr during the last 250 Myr. In order to explain this 26 Myr periodicity, it was proposed that the mass extinctions were caused by periodic increases in cometary impacts. One method to test this hypothesis is to determine if there were periodic increases in impact events (based on crater ages) that correlate with mass extinctions. A way to test the hypothesis that mass extinctions were caused by periodic increases in impact cratering is to look for evidence of impact events in deep-sea deposits. This method allows direct observation of the temporal relationship between impact events and extinctions as recorded in the sedimentary record. There is evidence in the deep-sea record for two (possibly three) impact events in the late Eocene. The younger event, represented by the North American microtektite layer, is not associated with an Ir anomaly. The older event, defined by the cpx spherule layer, is associated with an Ir anomaly. However, neither of the two impact events recorded in late Eocene deposits appears to be associated with an unusual number of extinctions. Thus there is little evidence in the deep-sea record for an impact-related mass extinction in the late Eocene.

Deep multi-scale convolutional neural network for hyperspectral image classification

Science.gov (United States)

Zhang, Feng-zhe; Yang, Xia

2018-04-01

In this paper, we proposed a multi-scale convolutional neural network for hyperspectral image classification task. Firstly, compared with conventional convolution, we utilize multi-scale convolutions, which possess larger respective fields, to extract spectral features of hyperspectral image. We design a deep neural network with a multi-scale convolution layer which contains 3 different convolution kernel sizes. Secondly, to avoid overfitting of deep neural network, dropout is utilized, which randomly sleeps neurons, contributing to improve the classification accuracy a bit. In addition, new skills like ReLU in deep learning is utilized in this paper. We conduct experiments on University of Pavia and Salinas datasets, and obtained better classification accuracy compared with other methods.

Dilution limits dissolved organic carbon utilization in the deep ocean

KAUST Repository

Arrieta, Jesus

2015-03-19

Oceanic dissolved organic carbon (DOC) is the second largest reservoir of organic carbon in the biosphere. About 72% of the global DOC inventory is stored in deep oceanic layers for years to centuries, supporting the current view that it consists of materials resistant to microbial degradation. An alternative hypothesis is that deep-water DOC consists of many different, intrinsically labile compounds at concentrations too low to compensate for the metabolic costs associated to their utilization. Here, we present experimental evidence showing that low concentrations rather than recalcitrance preclude consumption of a substantial fraction of DOC, leading to slow microbial growth in the deep ocean. These findings demonstrate an alternative mechanism for the long-term storage of labile DOC in the deep ocean, which has been hitherto largely ignored. © 2015, American Association for the Advancement of Science. All rights reserved.

Dilution limits dissolved organic carbon utilization in the deep ocean

KAUST Repository

Arrieta, J M; Mayol, Eva; Hansman, Roberta L.; Herndl, Gerhard J.; Dittmar, Thorsten; Duarte, Carlos M.

2015-01-01

Oceanic dissolved organic carbon (DOC) is the second largest reservoir of organic carbon in the biosphere. About 72% of the global DOC inventory is stored in deep oceanic layers for years to centuries, supporting the current view that it consists of materials resistant to microbial degradation. An alternative hypothesis is that deep-water DOC consists of many different, intrinsically labile compounds at concentrations too low to compensate for the metabolic costs associated to their utilization. Here, we present experimental evidence showing that low concentrations rather than recalcitrance preclude consumption of a substantial fraction of DOC, leading to slow microbial growth in the deep ocean. These findings demonstrate an alternative mechanism for the long-term storage of labile DOC in the deep ocean, which has been hitherto largely ignored. © 2015, American Association for the Advancement of Science. All rights reserved.

Integrating deep and shallow natural language processing components : representations and hybrid architectures

OpenAIRE

Schäfer, Ulrich

2006-01-01

We describe basic concepts and software architectures for the integration of shallow and deep (linguistics-based, semantics-oriented) natural language processing (NLP) components. The main goal of this novel, hybrid integration paradigm is improving robustness of deep processing. After an introduction to constraint-based natural language parsing, we give an overview of typical shallow processing tasks. We introduce XML standoff markup as an additional abstraction layer that eases integration ...

Moving object detection in video satellite image based on deep learning

Science.gov (United States)

Zhang, Xueyang; Xiang, Junhua

2017-11-01

Moving object detection in video satellite image is studied. A detection algorithm based on deep learning is proposed. The small scale characteristics of remote sensing video objects are analyzed. Firstly, background subtraction algorithm of adaptive Gauss mixture model is used to generate region proposals. Then the objects in region proposals are classified via the deep convolutional neural network. Thus moving objects of interest are detected combined with prior information of sub-satellite point. The deep convolution neural network employs a 21-layer residual convolutional neural network, and trains the network parameters by transfer learning. Experimental results about video from Tiantuo-2 satellite demonstrate the effectiveness of the algorithm.

Model-free prediction of noisy chaotic time series by deep learning

OpenAIRE

Yeo, Kyongmin

2017-01-01

We present a deep neural network for a model-free prediction of a chaotic dynamical system from noisy observations. The proposed deep learning model aims to predict the conditional probability distribution of a state variable. The Long Short-Term Memory network (LSTM) is employed to model the nonlinear dynamics and a softmax layer is used to approximate a probability distribution. The LSTM model is trained by minimizing a regularized cross-entropy function. The LSTM model is validated against...

Deep learning-based Diabetic Retinopathy assessment on embedded system.

Science.gov (United States)

Ardiyanto, Igi; Nugroho, Hanung Adi; Buana, Ratna Lestari Budiani

2017-07-01

Diabetic Retinopathy (DR) is a disease which affect the vision ability. The observation by an ophthalmologist usually conducted by analyzing the retinal images of the patient which are marked by some DR features. However some misdiagnosis are usually found due to human error. Here, a deep learning-based low-cost embedded system is established to assist the doctor for grading the severity of the DR from the retinal images. A compact deep learning algorithm named Deep-DR-Net which fits on a small embedded board is afterwards proposed for such purposes. In the heart of Deep-DR-Net, a cascaded encoder-classifier network is arranged using residual style for ensuring the small model size. The usage of different types of convolutional layers subsequently guarantees the features richness of the network for differentiating the grade of the DR. Experimental results show the capability of the proposed system for detecting the existence as well as grading the severity of the DR symptomps.

DeepNet: An Ultrafast Neural Learning Code for Seismic Imaging

International Nuclear Information System (INIS)

Barhen, J.; Protopopescu, V.; Reister, D.

1999-01-01

A feed-forward multilayer neural net is trained to learn the correspondence between seismic data and well logs. The introduction of a virtual input layer, connected to the nominal input layer through a special nonlinear transfer function, enables ultrafast (single iteration), near-optimal training of the net using numerical algebraic techniques. A unique computer code, named DeepNet, has been developed, that has achieved, in actual field demonstrations, results unattainable to date with industry standard tools

Radio-active waste disposal and deep-sea biology

International Nuclear Information System (INIS)

Rice, A.L.

1978-01-01

The deep-sea has been widely thought of as a remote, sparsely populated, and biologically inactive environment, well suited to receive the noxious products of nuclear fission processes. Much of what is known of abyssal biology tends to support this view, but there are a few disquieting contra-indications. The realisation, in recent years, that many animal groups show a previously unsuspected high species diversity in the deep-sea emphasized the paucity of our knowledge of this environment. More dramatically, the discovery of a large, active, and highly mobile abysso-bentho-pelagic fauna changed the whole concept of abyssal life. Finally, while there is little evidence for the existence of vertical migration patterns linking the deep-sea bottom communities with those of the overlying water layers, there are similarly too few negative results for the possibility of such transport mechanisms to be dismissed. In summary, biological knowledge of the abyss is insufficient to answer the questions raised in connection with deep-sea dumping, but in the absence of adequate answers it might be dangerous to ignore the questions

A Deep Learning Approach for Fault Diagnosis of Induction Motors in Manufacturing

Science.gov (United States)

Shao, Si-Yu; Sun, Wen-Jun; Yan, Ru-Qiang; Wang, Peng; Gao, Robert X.

2017-11-01

Extracting features from original signals is a key procedure for traditional fault diagnosis of induction motors, as it directly influences the performance of fault recognition. However, high quality features need expert knowledge and human intervention. In this paper, a deep learning approach based on deep belief networks (DBN) is developed to learn features from frequency distribution of vibration signals with the purpose of characterizing working status of induction motors. It combines feature extraction procedure with classification task together to achieve automated and intelligent fault diagnosis. The DBN model is built by stacking multiple-units of restricted Boltzmann machine (RBM), and is trained using layer-by-layer pre-training algorithm. Compared with traditional diagnostic approaches where feature extraction is needed, the presented approach has the ability of learning hierarchical representations, which are suitable for fault classification, directly from frequency distribution of the measurement data. The structure of the DBN model is investigated as the scale and depth of the DBN architecture directly affect its classification performance. Experimental study conducted on a machine fault simulator verifies the effectiveness of the deep learning approach for fault diagnosis of induction motors. This research proposes an intelligent diagnosis method for induction motor which utilizes deep learning model to automatically learn features from sensor data and realize working status recognition.

Multi-level deep supervised networks for retinal vessel segmentation.

Science.gov (United States)

Mo, Juan; Zhang, Lei

2017-12-01

Changes in the appearance of retinal blood vessels are an important indicator for various ophthalmologic and cardiovascular diseases, including diabetes, hypertension, arteriosclerosis, and choroidal neovascularization. Vessel segmentation from retinal images is very challenging because of low blood vessel contrast, intricate vessel topology, and the presence of pathologies such as microaneurysms and hemorrhages. To overcome these challenges, we propose a neural network-based method for vessel segmentation. A deep supervised fully convolutional network is developed by leveraging multi-level hierarchical features of the deep networks. To improve the discriminative capability of features in lower layers of the deep network and guide the gradient back propagation to overcome gradient vanishing, deep supervision with auxiliary classifiers is incorporated in some intermediate layers of the network. Moreover, the transferred knowledge learned from other domains is used to alleviate the issue of insufficient medical training data. The proposed approach does not rely on hand-crafted features and needs no problem-specific preprocessing or postprocessing, which reduces the impact of subjective factors. We evaluate the proposed method on three publicly available databases, the DRIVE, STARE, and CHASE_DB1 databases. Extensive experiments demonstrate that our approach achieves better or comparable performance to state-of-the-art methods with a much faster processing speed, making it suitable for real-world clinical applications. The results of cross-training experiments demonstrate its robustness with respect to the training set. The proposed approach segments retinal vessels accurately with a much faster processing speed and can be easily applied to other biomedical segmentation tasks.

Deep-well injection of radioactive waste in Russia

International Nuclear Information System (INIS)

Hoek, J.

1998-01-01

In the Russian federation, deep borehole injection of liquid radioactive waste has been established practice since at least 1963. The liquid is injected into sandy or other formations with high porosity, which are isolated by water-tight layers. This technique has also been used elsewhere for toxic liquid waste and residues from mining operations. Deep-well injection of radioactive waste is not currently used in any of the European Commission (EC) countries. In this paper the results of a EC-funded study were presented. The study is entitled 'Measurements, modelling of migration and possible radiological consequences at deep well injection sites for liquid radioactive waste in Russia', COSU-CT94-0099-UK. The study was carried out jointly by AEA Technology, CAG and the Research Institute for Nuclear Reactors (NIIAR) at Dimitrovgrad. Many scientists have contributed to the results reported here. The aims of the study are: Provision of extensive information on the deep-well injection repositories and their use in the former Soviet Union; Provision of a methodology to assess safety aspects of deep-well injection of liquid radioactive waste in deep geological formations; This will allow evaluation of proposals to use deep-well injection techniques in other regions; Support for Russian regulatory bodies through evaluation of the suitability of the sites, including estimates of the maximum amount of waste that can be safely stored in them; and Provision of a methodology to assess the use of deep-well injection repositories as an alternative disposal technique for EC countries. 7 refs

E-beam irradiation effect on CdSe/ZnSe QD formation by MBE: deep level transient spectroscopy and cathodoluminescence studies

International Nuclear Information System (INIS)

Kozlovsky, V I; Litvinov, V G; Sadofyev, Yu G

2004-01-01

CdSe/ZnSe structures containing 1 or 15 thin (3-5 monolayers) CdSe layers were studied by cathodoluminescence (CL) and deep level transient spectroscopy (DLTS). The DLTS spectra consisted of peaks from deep levels (DLs) and an additional intense peak due to electron emission from the ground quantized level in the CdSe layers. Activation energy of this additional peak correlated with an energy of the CdSe-layer emission line in the CL spectra. Electron-beam irradiation of the structure during the growth process was found to influence the DLTS and CL spectra of the CdSe layers, shifting the CdSe-layer emission line to the long-wave side. The obtained results are explained using the assumption that e-beam irradiation stimulates the formation of quantum dots of various sizes in the CdSe layers

Magnetooptic effects and Auger electron spectroscopy of two-layer NiFe-Dy and Fe-Dy films with nonuniform layers

International Nuclear Information System (INIS)

Ehdel'man, I.S.; Markov, V.V.; Khudyakov, A.E.; Ivantsov, R.D.; Bondarenko, G.V.; Ovchinnikov, S.G.; Kesler, V.G.; Parshin, A.S.; Ronzhin, I.P.

2001-01-01

Magneto-optical effects (magnetic circular dichroism and meridional Kerr effect) and element distribution with layer thickness in two-layer NiFe-Dy and Fe-Dy films, prepared by thermal sputtering of component in ultrahigh vacuum, are investigated. It is shown, that Dy in a two-layer film in the temperature range of 80-300 K makes constant contributions to both effects investigated which are approximately equal to the values of the effects observed in an isolated Dy film only at temperatures below the temperature T c of Dy transition into a ferromagnetic state (T c ∼ 100 K for the films under study). This behaviour of magneto-optical effects is assumed to be due to the influence of a NiFe layer spin system on magnetic state of a Dy layer, this influence is enhanced by the deep penetration of Ni and Fe ions into Dy layer as it follows from the data obtained using Auger electron spectroscopy [ru

Induction of superficial cortical layer neurons from mouse embryonic stem cells by valproic acid.

Science.gov (United States)

Juliandi, Berry; Abematsu, Masahiko; Sanosaka, Tsukasa; Tsujimura, Keita; Smith, Austin; Nakashima, Kinichi

2012-01-01

Within the developing mammalian cortex, neural progenitors first generate deep-layer neurons and subsequently more superficial-layer neurons, in an inside-out manner. It has been reported recently that mouse embryonic stem cells (mESCs) can, to some extent, recapitulate cortical development in vitro, with the sequential appearance of neurogenesis markers resembling that in the developing cortex. However, mESCs can only recapitulate early corticogenesis; superficial-layer neurons, which are normally produced in later developmental periods in vivo, are under-represented. This failure of mESCs to reproduce later corticogenesis in vitro implies the existence of crucial factor(s) that are absent or uninduced in existing culture systems. Here we show that mESCs can give rise to superficial-layer neurons efficiently when treated with valproic acid (VPA), a histone deacetylase inhibitor. VPA treatment increased the production of Cux1-positive superficial-layer neurons, and decreased that of Ctip2-positive deep-layer neurons. These results shed new light on the mechanisms of later corticogenesis. Copyright © 2011 Elsevier Ireland Ltd and the Japan Neuroscience Society. All rights reserved.

The biomass of the deep-sea benthopelagic plankton

Science.gov (United States)

Wishner, K. F.

1980-04-01

Deep-sea benthopelagic plankton samples were collected with a specially designed opening-closing net system 10 to 100 m above the bottom in five different oceanic regions at depths from 1000 to 4700 m. Benthopelagic plankton biomasses decrease exponentially with depth. At 1000 m the biomass is about 1% that of the surface zooplankton, at 5000 m about 0.1%. Effects of differences in surface primary productivity on deep-sea plankton biomass are much less than the effect of depth and are detectable only in a few comparisons of extreme oceanic regions. The biomass at 10 m above the bottom is greater than that at 100 m above the bottom (in a three-sample comparison), which could be a consequence of an enriched near-bottom environment. The deep-sea plankton biomass in the Red Sea is anomalously low. This may be due to increased decomposition rates in the warm (22°C) deep Red Sea water, which prevent much detritus from reaching the deep sea. A model of organic carbon utilization in the benthic boundary layer (bottom 100 m), incorporating results from deep-sea sediment trap and respiration studies, indicates that the benthopelagic plankton use only a small amount of the organic carbon flux. A large fraction of the flux is unaccounted for by present estimates of benthic and benthopelagic respiration.

A deep belief network with PLSR for nonlinear system modeling.

Science.gov (United States)

Qiao, Junfei; Wang, Gongming; Li, Wenjing; Li, Xiaoli

2017-10-31

Nonlinear system modeling plays an important role in practical engineering, and deep learning-based deep belief network (DBN) is now popular in nonlinear system modeling and identification because of the strong learning ability. However, the existing weights optimization for DBN is based on gradient, which always leads to a local optimum and a poor training result. In this paper, a DBN with partial least square regression (PLSR-DBN) is proposed for nonlinear system modeling, which focuses on the problem of weights optimization for DBN using PLSR. Firstly, unsupervised contrastive divergence (CD) algorithm is used in weights initialization. Secondly, initial weights derived from CD algorithm are optimized through layer-by-layer PLSR modeling from top layer to bottom layer. Instead of gradient method, PLSR-DBN can determine the optimal weights using several PLSR models, so that a better performance of PLSR-DBN is achieved. Then, the analysis of convergence is theoretically given to guarantee the effectiveness of the proposed PLSR-DBN model. Finally, the proposed PLSR-DBN is tested on two benchmark nonlinear systems and an actual wastewater treatment system as well as a handwritten digit recognition (nonlinear mapping and modeling) with high-dimension input data. The experiment results show that the proposed PLSR-DBN has better performances of time and accuracy on nonlinear system modeling than that of other methods. Copyright © 2017 Elsevier Ltd. All rights reserved.

Deep Ocean Contribution to Sea Level Rise

Science.gov (United States)

Chang, L.; Sun, W.; Tang, H.; Wang, Q.

2017-12-01

The ocean temperature and salinity change in the upper 2000m can be detected by Argo floats, so we can know the steric height change of the ocean. But the ocean layers above 2000m represent only 50% of the total ocean volume. Although the temperature and salinity change are small compared to the upper ocean, the deep ocean contribution to sea level might be significant because of its large volume. There has been some research on the deep ocean rely on the very sparse situ observation and are limited to decadal and longer-term rates of change. The available observational data in the deep ocean are too spares to determine the temporal variability, and the long-term changes may have a bias. We will use the Argo date and combine the situ data and topographic data to estimate the temperature and salinity of the sea water below 2000m, so we can obtain a monthly data. We will analyze the seasonal and annual change of the steric height change due to the deep ocean between 2005 and 2016. And we will evaluate the result combination the present-day satellite and in situ observing systems. The deep ocean contribution can be inferred indirectly as the difference between the altimetry minus GRACE and Argo-based steric sea level.

DeepPy: Pythonic deep learning

DEFF Research Database (Denmark)

Larsen, Anders Boesen Lindbo

This technical report introduces DeepPy – a deep learning framework built on top of NumPy with GPU acceleration. DeepPy bridges the gap between highperformance neural networks and the ease of development from Python/NumPy. Users with a background in scientific computing in Python will quickly...... be able to understand and change the DeepPy codebase as it is mainly implemented using high-level NumPy primitives. Moreover, DeepPy supports complex network architectures by letting the user compose mathematical expressions as directed graphs. The latest version is available at http...

Formation of conductive spontaneous via holes in AlN buffer layer on n+Si substrate by filling the vias with n-AlGaN by metal organic chemical vapor deposition and application to vertical deep ultraviolet photo-sensor

Directory of Open Access Journals (Sweden)

N. Kurose

2014-12-01

Full Text Available We have grown conductive aluminum nitride (AlN layers using the spontaneous via holes formation technique on an n+-Si substrate for vertical-type device fabrication. The size and density of the via holes are controlled through the crystal growth conditions used for the layer, and this enables the conductance of the layer to be controlled. Using this technique, we demonstrate the fabrication of a vertical-type deep ultraviolet (DUV photo-sensor. This technique opens up the possibility of fabrication of monolithically integrated on-chip DUV sensors and DUV light-emitting devices (LEDs, including amplifiers, controllers and other necessary functional circuits, on a Si substrate.

The circulation of deep water in the Tasman and Coral seas

International Nuclear Information System (INIS)

Harries, J.R.

1976-07-01

The physical oceanography of the Tasman and Coral Seas is reviewed with an emphasis on the deep currents. There are many uncertainties in the deep circulation pattern. The available data are used to develop an idealised circulation to estimate the likely path taken by water flowing from a depth of 5000 m in the Tasman Sea. The model suggests that the water would finally reach the surface layers south of the Antarctic Convergence with a median delay of 600 years. (author)

Bottom water circulation in Cascadia Basin

Science.gov (United States)

Hautala, Susan L.; Paul Johnson, H.; Hammond, Douglas E.

2009-10-01

A combination of beta spiral and minimum length inverse methods, along with a compilation of historical and recent high-resolution CTD data, are used to produce a quantitative estimate of the subthermocline circulation in Cascadia Basin. Flow in the North Pacific Deep Water, from 900-1900 m, is characterized by a basin-scale anticyclonic gyre. Below 2000 m, two water masses are present within the basin interior, distinguished by different potential temperature-salinity lines. These water masses, referred to as Cascadia Basin Bottom Water (CBBW) and Cascadia Basin Deep Water (CBDW), are separated by a transition zone at about 2400 m depth. Below the depth where it freely communicates with the broader North Pacific, Cascadia Basin is renewed by northward flow through deep gaps in the Blanco Fracture Zone that feeds the lower limb of a vertical circulation cell within the CBBW. Lower CBBW gradually warms and returns to the south at lighter density. Isopycnal layer renewal times, based on combined lateral and diapycnal advective fluxes, increase upwards from the bottom. The densest layer, existing in the southeast quadrant of the basin below ˜2850 m, has an advective flushing time of 0.6 years. The total volume flushing time for the entire CBBW is 2.4 years, corresponding to an average water parcel residence time of 4.7 years. Geothermal heating at the Cascadia Basin seafloor produces a characteristic bottom-intensified temperature anomaly and plays an important role in the conversion of cold bottom water to lighter density within the CBBW. Although covering only about 0.05% of the global seafloor, the combined effects of bottom heat flux and diapycnal mixing within Cascadia Basin provide about 2-3% of the total required global input to the upward branch of the global thermohaline circulation.

A Spatiotemporal Prediction Framework for Air Pollution Based on Deep RNN

Directory of Open Access Journals (Sweden)

J. Fan

2017-10-01

Full Text Available Time series data in practical applications always contain missing values due to sensor malfunction, network failure, outliers etc. In order to handle missing values in time series, as well as the lack of considering temporal properties in machine learning models, we propose a spatiotemporal prediction framework based on missing value processing algorithms and deep recurrent neural network (DRNN. By using missing tag and missing interval to represent time series patterns, we implement three different missing value fixing algorithms, which are further incorporated into deep neural network that consists of LSTM (Long Short-term Memory layers and fully connected layers. Real-world air quality and meteorological datasets (Jingjinji area, China are used for model training and testing. Deep feed forward neural networks (DFNN and gradient boosting decision trees (GBDT are trained as baseline models against the proposed DRNN. Performances of three missing value fixing algorithms, as well as different machine learning models are evaluated and analysed. Experiments show that the proposed DRNN framework outperforms both DFNN and GBDT, therefore validating the capacity of the proposed framework. Our results also provides useful insights for better understanding of different strategies that handle missing values.

Seismic response analysis of the deep saturated soil deposits in Shanghai

Science.gov (United States)

Huang, Yu; Ye, Weimin; Chen, Zhuchang

2009-01-01

The quaternary deposits in Shanghai are horizontal soil layers of thickness up to about 280 m in the urban area with an annual groundwater table between 0.5 and 0.7 m from the surface. The characteristics of deep saturated deposits may have important influences upon seismic response of the ground in Shanghai. Based on the Biot theory for porous media, the water-saturated soil deposits are modeled as a two-phase porous system consisting of solid and fluid phases, in this paper. A nonlinear constitutive model for predicting the seismic response of the ground is developed to describe the dynamic characters of the deep-saturated soil deposits in Shanghai. Subsequently, the seismic response of a typical site with 280 m deep soil layers, which is subjected to four base excitations (El Centro, Taft, Sunan, and Tangshan earthquakes), is analyzed in terms of an effective stress-based finite element method with the proposed constitutive model. Special emphasis is given to the computed results of accelerations, excess pore-water pressures, and settlements during the seismic excitations. It has been found that the analysis can capture fundamental aspects of the ground response and produce preliminary results for seismic assessment.

A Spatiotemporal Prediction Framework for Air Pollution Based on Deep RNN

Science.gov (United States)

Fan, J.; Li, Q.; Hou, J.; Feng, X.; Karimian, H.; Lin, S.

2017-10-01

Time series data in practical applications always contain missing values due to sensor malfunction, network failure, outliers etc. In order to handle missing values in time series, as well as the lack of considering temporal properties in machine learning models, we propose a spatiotemporal prediction framework based on missing value processing algorithms and deep recurrent neural network (DRNN). By using missing tag and missing interval to represent time series patterns, we implement three different missing value fixing algorithms, which are further incorporated into deep neural network that consists of LSTM (Long Short-term Memory) layers and fully connected layers. Real-world air quality and meteorological datasets (Jingjinji area, China) are used for model training and testing. Deep feed forward neural networks (DFNN) and gradient boosting decision trees (GBDT) are trained as baseline models against the proposed DRNN. Performances of three missing value fixing algorithms, as well as different machine learning models are evaluated and analysed. Experiments show that the proposed DRNN framework outperforms both DFNN and GBDT, therefore validating the capacity of the proposed framework. Our results also provides useful insights for better understanding of different strategies that handle missing values.

Improving model biases in an ESM with an isopycnic ocean component by accounting for wind work on oceanic near-inertial motions.

Science.gov (United States)

de Wet, P. D.; Bentsen, M.; Bethke, I.

2016-02-01

It is well-known that, when comparing climatological parameters such as ocean temperature and salinity to the output of an Earth System Model (ESM), the model exhibits biases. In ESMs with an isopycnic ocean component, such as NorESM, insufficient vertical mixing is thought to be one of the causes of such differences between observational and model data. However, enhancing the vertical mixing of the model's ocean component not only requires increasing the energy input, but also sound physical reasoning for doing so. Various authors have shown that the action of atmospheric winds on the ocean's surface is a major source of energy input into the upper ocean. However, due to model and computational constraints, oceanic processes linked to surface winds are incompletely accounted for. Consequently, despite significantly contributing to the energy required to maintain ocean stratification, most ESMs do not directly make provision for this energy. In this study we investigate the implementation of a routine in which the energy from work done on oceanic near-inertial motions is calculated in an offline slab model. The slab model, which has been well-documented in the literature, runs parallel to but independently from the ESM's ocean component. It receives wind fields with a frequency higher than that of the coupling frequency, allowing it to capture the fluctuations in the winds on shorter time scales. The additional energy calculated thus is then passed to the ocean component, avoiding the need for increased coupling between the components of the ESM. Results show localised reduction in, amongst others, the salinity and temperature biases of NorESM, confirming model sensitivity to wind-forcing and points to the need for better representation of surface processes in ESMs.

Printed wax masks for 254 nm deep-UV pattering of PMMA-based microfluidics

International Nuclear Information System (INIS)

Fan, Yiqiang; Liu, Yang; Li, Huawei; Foulds, Ian G

2012-01-01

This paper reports a new technique for masking deep-UV exposure of poly(methyl methacrylate) (PMMA) using a printed wax mask. This technique provides an inexpensive and bulk fabrication method for PMMA structures. The technique involves the direct printing of the mask onto a polymer sheet using a commercial wax printer. The wax layer was then transferred to a PMMA substrate using a thermal laminator, exposed using deep-UV (with a wavelength of 254 nm), developed in an IPA:water solution, and completed by bonding on a PMMA cap layer. A sample microfluidic device fabricated with this method is also presented, with the microchannel as narrow as 50 µm. The whole process is easy to perform without the requirement for any microfabrication facilities. (technical note)

Benthic boundary layer modelling studies

International Nuclear Information System (INIS)

Richards, K.J.

1984-01-01

A numerical model has been developed to study the factors which control the height of the benthic boundary layer in the deep ocean and the dispersion of a tracer within and directly above the layer. This report covers tracer clouds of horizontal scales of 10 to 100 km. The dispersion of a tracer has been studied in two ways. Firstly, a number of particles have been introduced into the flow. The trajectories of these particles provide information on dispersion rates. For flow conditions similar to those observed in the abyssal N.E. Atlantic the diffusivity of a tracer was found to be 5 x 10 6 cm 2 s -1 for a tracer within the boundary layer and 8 x 10 6 cm 2 s -1 for a tracer above the boundary layer. The results are in accord with estimates made from current meter measurements. The second method of studying dispersion was to calculate the evolution of individual tracer clouds. Clouds within and above the benthic boundary layer often show quite different behaviour from each other although the general structure of the clouds in the two regions were found to have no significant differences. (author)

Dilution limits dissolved organic carbon utilization in the deep ocean

NARCIS (Netherlands)

Arrieta, J.M.; Mayol, E.; Hansman, R.L.; Herndl, G.J.; Dittmar, T.; Duarte, C.M.

2015-01-01

Oceanic dissolved organic carbon (DOC) is the second largest reservoir of organic carbon in the biosphere. About 72% of the global DOC inventory is stored in deep oceanic layers for years to centuries, supporting the current view that it consists of materials resistant to microbial degradation. An

Laser micromachined wax-covered plastic paper as both sputter deposition shadow masks and deep-ultraviolet patterning masks for polymethylmethacrylate-based microfluidic systems

KAUST Repository

Fan, Yiqiang

2013-12-16

We report a technically innovative method of fabricating masks for both deep-ultraviolet (UV) patterning and metal sputtering on polymethylmethacrylate (PMMA) for microfluidic systems. We used a CO2 laser system to cut the required patterns on wax-covered plastic paper; the laser-patterned wax paper will either work as a mask for deep-UV patterning or as a mask for metal sputtering. A microfluidic device was also fabricated to demonstrate the feasibility of this method. The device has two layers: the first layer is a 1-mm thick PMMA substrate that was patterned by deep-UV exposure to create microchannels. The mask used in this process was the laser-cut wax paper. The second layer, also a 1-mm thick PMMA layer, was gold sputtered with patterned wax paper as the shadow mask. These two pieces of PMMA were then bonded to form microchannels with exposed electrodes. This process is a simple and rapid method for creating integrated microfluidic systems that do not require cleanroom facilities.

Biogeochemical characteristics of suspended particulate matter in deep chlorophyll maximum layers in the southern East China Sea

Science.gov (United States)

Liu, Qianqian; Kandasamy, Selvaraj; Lin, Baozhi; Wang, Huawei; Chen, Chen-Tung Arthur

2018-04-01

Continental shelves and marginal seas are key sites of particulate organic matter (POM) production, remineralization and sequestration, playing an important role in the global carbon cycle. Elemental and stable isotopic compositions of organic carbon and nitrogen are thus frequently used to characterize and distinguish POM and its sources in suspended particles and surface sediments in the marginal seas. Here we investigated suspended particulate matter (SPM) collected around deep chlorophyll maximum (DCM) layers in the southern East China Sea for particulate organic carbon and nitrogen (POC and PN) contents and their isotopic compositions (δ13CPOC and δ15NPN) to understand provenance and dynamics of POM. Hydrographic parameters (temperature, salinity and turbidity) indicated that the study area was weakly influenced by freshwater derived from the Yangtze River during summer 2013. Elemental and isotopic results showed a large variation in δ13CPOC (-25.8 to -18.2 ‰) and δ15NPN (3.8 to 8.0 ‰), but a narrow molar C / N ratio (4.1-6.3) and low POC / Chl a ratio ( < 200 g g-1) in POM, and indicated that the POM in DCM layers was newly produced by phytoplankton. In addition to temperature effects, the range and distribution of δ13CPOC were controlled by variations in primary productivity and phytoplankton species composition; the former explained ˜ 70 % of the variability in δ13CPOC. However, the variation in δ15NPN was controlled by the nutrient status and δ15NNO3- in seawater, as indicated by similar spatial distribution between δ15NPN and the current pattern and water masses in the East China Sea; although interpretations of δ15NPN data should be verified with the nutrient data in future studies. Furthermore, the POM investigated was weakly influenced by the terrestrial OM supplied by the Yangtze River during summer 2013 due to the reduced sediment supply by the Yangtze River and north-eastward transport of riverine particles to the northern East China

Improving the Separability of Deep Features with Discriminative Convolution Filters for RSI Classification

Directory of Open Access Journals (Sweden)

Na Liu

2018-03-01

Full Text Available The extraction of activation vectors (or deep features from the fully connected layers of a convolutional neural network (CNN model is widely used for remote sensing image (RSI representation. In this study, we propose to learn discriminative convolution filter (DCF based on class-specific separability criteria for linear transformation of deep features. In particular, two types of pretrained CNN called CaffeNet and VGG-VD16 are introduced to illustrate the generality of the proposed DCF. The activation vectors extracted from the fully connected layers of a CNN are rearranged into the form of an image matrix, from which a spatial arrangement of local patches is extracted using sliding window strategy. DCF learning is then performed on each local patch individually to obtain the corresponding discriminative convolution kernel through generalized eigenvalue decomposition. The proposed DCF learning characterizes that a convolutional kernel with small size (e.g., 3 × 3 pixels can be effectively learned on a small-size local patch (e.g., 8 × 8 pixels, thereby ensuring that the linear transformation of deep features can maintain low computational complexity. Experiments on two RSI datasets demonstrate the effectiveness of DCF in improving the classification performances of deep features without increasing dimensionality.

Microbial diversity in methane hydrate-bearing deep marine sediments core preserved in the original pressure.

Science.gov (United States)

Takahashi, Y.; Hata, T.; Nishida, H.

2017-12-01

In normal coring of deep marine sediments, the sampled cores are exposed to the pressure of the atmosphere, which results in dissociation of gas-hydrates and might change microbial diversity. In this study, we analyzed microbial composition in methane hydrate-bearing sediment core sampled and preserved by Hybrid-PCS (Pressure Coring System). We sliced core into three layers; (i) outside layer, which were most affected by drilling fluids, (ii) middle layer, and (iii) inner layer, which were expected to be most preserved as the original state. From each layer, we directly extracted DNA, and amplified V3-V4 region of 16S rRNA gene. We determined at least 5000 of nucleotide sequences of the partial 16S rDNA from each layer by Miseq (Illumina). In the all layers, facultative anaerobes, which can grow with or without oxygen because they can metabolize energy aerobically or anaerobically, were detected as majority. However, the genera which are often detected anaerobic environment is abundant in the inner layer compared to the outside layer, indicating that condition of drilling and preservation affect the microbial composition in the deep marine sediment core. This study was conducted as a part of the activity of the Research Consortium for Methane Hydrate Resources in Japan [MH21 consortium], and supported by JOGMEC (Japan Oil, Gas and Metals National Corporation). The sample was provided by AIST (National Institute of Advanced Industrial Science and Technology).

Deep sequencing of ESTs from nacreous and prismatic layer producing tissues and a screen for novel shell formation-related genes in the pearl oyster.

Directory of Open Access Journals (Sweden)

Shigeharu Kinoshita

Full Text Available BACKGROUND: Despite its economic importance, we have a limited understanding of the molecular mechanisms underlying shell formation in pearl oysters, wherein the calcium carbonate crystals, nacre and prism, are formed in a highly controlled manner. We constructed comprehensive expressed gene profiles in the shell-forming tissues of the pearl oyster Pinctada fucata and identified novel shell formation-related genes candidates. PRINCIPAL FINDINGS: We employed the GS FLX 454 system and constructed transcriptome data sets from pallial mantle and pearl sac, which form the nacreous layer, and from the mantle edge, which forms the prismatic layer in P. fucata. We sequenced 260477 reads and obtained 29682 unique sequences. We also screened novel nacreous and prismatic gene candidates by a combined analysis of sequence and expression data sets, and identified various genes encoding lectin, protease, protease inhibitors, lysine-rich matrix protein, and secreting calcium-binding proteins. We also examined the expression of known nacreous and prismatic genes in our EST library and identified novel isoforms with tissue-specific expressions. CONCLUSIONS: We constructed EST data sets from the nacre- and prism-producing tissues in P. fucata and found 29682 unique sequences containing novel gene candidates for nacreous and prismatic layer formation. This is the first report of deep sequencing of ESTs in the shell-forming tissues of P. fucata and our data provide a powerful tool for a comprehensive understanding of the molecular mechanisms of molluscan biomineralization.

Isotope study on the Keuper sandstone aquifer with a leaky cover layer

International Nuclear Information System (INIS)

Geyh, M.A.; Backhaus, G.; Andres, G.; Rudolph, J.; Rath, H.K.

1984-01-01

Analyses of 14 C, 3 H, 39 Ar, delta 13 C and delta 18 O were performed on groundwater samples taken from the confined Keuper sandstone aquifer north of Nuremberg. The conventional 14 C data apparently contradict the hydrodynamic concept that the age of the deep groundwater flowing from east to west increases in the same direction. A two-dimensional dispersion model is used to convert the conventional 14 C groundwater ages to the regionally valid hydraulic conductivity coefficient of the leaky cover layer confining the aquifer. The basic assumption is that the deep groundwater has a water component which has percolated through the cover layer and which, on mixing, has changed the 14 C ages of the deep groundwater. Therefore, the ratio of the water from 'leaky' recharge to the water from the catchment area plays an important role. Values of delta 18 O and recharge temperatures derived from the noble-gas content of the deep water indicate mixing of Holocene and Pleistocene groundwaters and confirm the model. The considerable differences between the 39 Ar and 14 C groundwater ages may be plausibly explained by the hydrodynamic situation if 39 Ar production in the aquitard is assumed. (author)

Deep Joint Rain Detection and Removal from a Single Image

OpenAIRE

Yang, Wenhan; Tan, Robby T.; Feng, Jiashi; Liu, Jiaying; Guo, Zongming; Yan, Shuicheng

2016-01-01

In this paper, we address a rain removal problem from a single image, even in the presence of heavy rain and rain streak accumulation. Our core ideas lie in the new rain image models and a novel deep learning architecture. We first modify an existing model comprising a rain streak layer and a background layer, by adding a binary map that locates rain streak regions. Second, we create a new model consisting of a component representing rain streak accumulation (where individual streaks cannot b...

Electronic THz-spectrometer for plasmonic enhanced deep subwavelength layer detection

NARCIS (Netherlands)

Berrier, A.; Schaafsma, M.C.; Gómez Rivas, J.; Schäfer-Eberwein, H.; Haring Bolivar, P.; Tripodi, L.; Matters-Kammerer, M.K.

2015-01-01

We demonstrate the operation of a miniaturized all-electronic CMOS based THz spectrometer with performances comparable to that of a THz-TDS spectrometer in the frequency range 20 to 220 GHz. The use of this all-electronic THz spectrometer for detection of a thin TiO2 layer and a B. subtilis bacteria

Deep Structures of The Angola Margin

Science.gov (United States)

Moulin, M.; Contrucci, I.; Olivet, J.-L.; Aslanian, D.; Géli, L.; Sibuet, J.-C.

1 Ifremer Centre de Brest, DRO/Géosciences Marines, B.P. 70, 29280 Plouzané cedex (France) mmoulin@ifremer.fr/Fax : 33 2 98 22 45 49 2 Université de Bretagne Occidentale, Institut Universitaire Europeen de la Mer, Place Nicolas Copernic, 29280 Plouzane (France) 3 Total Fina Elf, DGEP/GSR/PN -GEOLOGIE, 2,place de la Coupole-La Defense 6, 92078 Paris la Defense Cedex Deep reflection and refraction seismic data were collected in April 2000 on the West African margin, offshore Angola, within the framework of the Zaiango Joint Project, conducted by Ifremer and Total Fina Elf Production. Vertical multichannel reflection seismic data generated by a « single-bubble » air gun array array (Avedik et al., 1993) were recorded on a 4.5 km long, digital streamer, while refraction and wide angle reflection seismic data were acquired on OBSs (Ocean Bottom Seismometers). Despite the complexity of the margin (5 s TWT of sediment, salt tectonics), the combination of seismic reflection and refraction methods results in an image and a velocity model of the ground structures below the Aptian salt layer. Three large seismic units appear in the reflection seismic section from the deep part on the margin under the base of salt. The upper seismic unit is layered with reflectors parallel to the base of the salt ; it represents unstructured sediments, filling a basin. The middle unit is seismically transparent. The lower unit is characterized by highly energetic reflectors. According to the OBS refraction data, these two units correspond to the continental crust and the base of the high energetic unit corresponds to the Moho. The margin appears to be divided in 3 domains, from east to west : i) a domain with an unthinned, 30 km thick, continental crust ; ii) a domain located between the hinge line and the foot of the continental slope, where the crust thins sharply, from 30 km to less than 7 km, this domain is underlain by an anormal layer with velocities comprising between 7,2 and 7

Growth temperature and dopant species effects on deep levels in Si grown by low temperature molecular beam epitaxy

International Nuclear Information System (INIS)

Chung, Sung-Yong; Jin, Niu; Rice, Anthony T.; Berger, Paul R.; Yu, Ronghua; Fang, Z-Q.; Thompson, Phillip E.

2003-01-01

Deep-level transient spectroscopy measurements were performed in order to investigate the effects of substrate growth temperature and dopant species on deep levels in Si layers during low-temperature molecular beam epitaxial growth. The structures studied were n + -p junctions using B doping for the p layer and p + -n junctions using P doping for the n layer. While the density of hole traps H1 (0.38-0.41 eV) in the B-doped p layers showed a clear increase with decreasing growth temperature from 600 to 370 degree sign C, the electron trap density was relatively constant. Interestingly, the minority carrier electron traps E1 (0.42-0.45 eV) and E2 (0.257 eV), found in the B-doped p layers, are similar to the majority carrier electron traps E11 (0.48 eV) and E22 (0.269 eV) observed in P-doped n layers grown at 600 degree sign C. It is hypothesized that these dominating electron traps are associated with pure divacancy defects and are independent of the dopant species

Understanding a Deep Learning Technique through a Neuromorphic System a Case Study with SpiNNaker Neuromorphic Platform

OpenAIRE

Sugiarto Indar; Pasila Felix

2018-01-01

Deep learning (DL) has been considered as a breakthrough technique in the field of artificial intelligence and machine learning. Conceptually, it relies on a many-layer network that exhibits a hierarchically non-linear processing capability. Some DL architectures such as deep neural networks, deep belief networks and recurrent neural networks have been developed and applied to many fields with incredible results, even comparable to human intelligence. However, many researchers are still scept...

Deep Convolutional Neural Networks: Structure, Feature Extraction and Training

Directory of Open Access Journals (Sweden)

Namatēvs Ivars

2017-12-01

Full Text Available Deep convolutional neural networks (CNNs are aimed at processing data that have a known network like topology. They are widely used to recognise objects in images and diagnose patterns in time series data as well as in sensor data classification. The aim of the paper is to present theoretical and practical aspects of deep CNNs in terms of convolution operation, typical layers and basic methods to be used for training and learning. Some practical applications are included for signal and image classification. Finally, the present paper describes the proposed block structure of CNN for classifying crucial features from 3D sensor data.

Bacterial niche-specific genome expansion is coupled with highly frequent gene disruptions in deep-sea sediments

KAUST Repository

Wang, Yong; Yang, Jiang Ke; Lee, On On; Li, Tie Gang; Al-Suwailem, Abdulaziz M.; Danchin, Antoine; Qian, Pei-Yuan

2011-01-01

The complexity and dynamics of microbial metagenomes may be evaluated by genome size, gene duplication and the disruption rate between lineages. In this study, we pyrosequenced the metagenomes of microbes obtained from the brine and sediment of a deep-sea brine pool in the Red Sea to explore the possible genomic adaptations of the microbes in response to environmental changes. The microbes from the brine and sediments (both surface and deep layers) of the Atlantis II Deep brine pool had similar communities whereas the effective genome size varied from 7.4 Mb in the brine to more than 9 Mb in the sediment. This genome expansion in the sediment samples was due to gene duplication as evidenced by enrichment of the homologs. The duplicated genes were highly disrupted, on average by 47.6% and 70% for the surface and deep layers of the Atlantis II Deep sediment samples, respectively. The disruptive effects appeared to be mainly due to point mutations and frameshifts. In contrast, the homologs from the Atlantis II Deep brine sample were highly conserved and they maintained relatively small copy numbers. Likely, the adaptation of the microbes in the sediments was coupled with pseudogenizations and possibly functional diversifications of the paralogs in the expanded genomes. The maintenance of the pseudogenes in the large genomes is discussed. © 2011 Wang et al.

Bacterial niche-specific genome expansion is coupled with highly frequent gene disruptions in deep-sea sediments

KAUST Repository

Wang, Yong

2011-12-21

The complexity and dynamics of microbial metagenomes may be evaluated by genome size, gene duplication and the disruption rate between lineages. In this study, we pyrosequenced the metagenomes of microbes obtained from the brine and sediment of a deep-sea brine pool in the Red Sea to explore the possible genomic adaptations of the microbes in response to environmental changes. The microbes from the brine and sediments (both surface and deep layers) of the Atlantis II Deep brine pool had similar communities whereas the effective genome size varied from 7.4 Mb in the brine to more than 9 Mb in the sediment. This genome expansion in the sediment samples was due to gene duplication as evidenced by enrichment of the homologs. The duplicated genes were highly disrupted, on average by 47.6% and 70% for the surface and deep layers of the Atlantis II Deep sediment samples, respectively. The disruptive effects appeared to be mainly due to point mutations and frameshifts. In contrast, the homologs from the Atlantis II Deep brine sample were highly conserved and they maintained relatively small copy numbers. Likely, the adaptation of the microbes in the sediments was coupled with pseudogenizations and possibly functional diversifications of the paralogs in the expanded genomes. The maintenance of the pseudogenes in the large genomes is discussed. © 2011 Wang et al.

Bacterial niche-specific genome expansion is coupled with highly frequent gene disruptions in deep-sea sediments.

Directory of Open Access Journals (Sweden)

Yong Wang

Full Text Available The complexity and dynamics of microbial metagenomes may be evaluated by genome size, gene duplication and the disruption rate between lineages. In this study, we pyrosequenced the metagenomes of microbes obtained from the brine and sediment of a deep-sea brine pool in the Red Sea to explore the possible genomic adaptations of the microbes in response to environmental changes. The microbes from the brine and sediments (both surface and deep layers of the Atlantis II Deep brine pool had similar communities whereas the effective genome size varied from 7.4 Mb in the brine to more than 9 Mb in the sediment. This genome expansion in the sediment samples was due to gene duplication as evidenced by enrichment of the homologs. The duplicated genes were highly disrupted, on average by 47.6% and 70% for the surface and deep layers of the Atlantis II Deep sediment samples, respectively. The disruptive effects appeared to be mainly due to point mutations and frameshifts. In contrast, the homologs from the Atlantis II Deep brine sample were highly conserved and they maintained relatively small copy numbers. Likely, the adaptation of the microbes in the sediments was coupled with pseudogenizations and possibly functional diversifications of the paralogs in the expanded genomes. The maintenance of the pseudogenes in the large genomes is discussed.

Fault Diagnosis Based on Chemical Sensor Data with an Active Deep Neural Network.

Science.gov (United States)

Jiang, Peng; Hu, Zhixin; Liu, Jun; Yu, Shanen; Wu, Feng

2016-10-13

Big sensor data provide significant potential for chemical fault diagnosis, which involves the baseline values of security, stability and reliability in chemical processes. A deep neural network (DNN) with novel active learning for inducing chemical fault diagnosis is presented in this study. It is a method using large amount of chemical sensor data, which is a combination of deep learning and active learning criterion to target the difficulty of consecutive fault diagnosis. DNN with deep architectures, instead of shallow ones, could be developed through deep learning to learn a suitable feature representation from raw sensor data in an unsupervised manner using stacked denoising auto-encoder (SDAE) and work through a layer-by-layer successive learning process. The features are added to the top Softmax regression layer to construct the discriminative fault characteristics for diagnosis in a supervised manner. Considering the expensive and time consuming labeling of sensor data in chemical applications, in contrast to the available methods, we employ a novel active learning criterion for the particularity of chemical processes, which is a combination of Best vs. Second Best criterion (BvSB) and a Lowest False Positive criterion (LFP), for further fine-tuning of diagnosis model in an active manner rather than passive manner. That is, we allow models to rank the most informative sensor data to be labeled for updating the DNN parameters during the interaction phase. The effectiveness of the proposed method is validated in two well-known industrial datasets. Results indicate that the proposed method can obtain superior diagnosis accuracy and provide significant performance improvement in accuracy and false positive rate with less labeled chemical sensor data by further active learning compared with existing methods.

A novel non-sequential hydrogen-pulsed deep reactive ion etching of silicon

International Nuclear Information System (INIS)

Gharooni, M; Mohajerzadeh, A; Sandoughsaz, A; Khanof, S; Mohajerzadeh, S; Asl-Soleimani, E

2013-01-01

A non-sequential pulsed-mode deep reactive ion etching of silicon is reported that employs continuous etching and passivation based on SF 6 and H 2 gases. The passivation layer, as an important step for deep vertical etching of silicon, is feasible by hydrogen pulses in proper time-slots. By adjusting the etching parameters such as plasma power, H 2 and SF 6 flows and hydrogen pulse timing, the process can be controlled for minimum underetch and high etch-rate at the same time. High-aspect-ratio features can be realized with low-density plasma power and by controlling the reaction chemistry. The so-called reactive ion etching lag has been minimized by operating the reactor at higher pressures. X-ray photoelectron spectroscopy and scanning electron microscopy have been used to study the formation of the passivation layer and the passivation mechanism. (paper)

Deep generative learning of location-invariant visual word recognition

Science.gov (United States)

Di Bono, Maria Grazia; Zorzi, Marco

2013-01-01

It is widely believed that orthographic processing implies an approximate, flexible coding of letter position, as shown by relative-position and transposition priming effects in visual word recognition. These findings have inspired alternative proposals about the representation of letter position, ranging from noisy coding across the ordinal positions to relative position coding based on open bigrams. This debate can be cast within the broader problem of learning location-invariant representations of written words, that is, a coding scheme abstracting the identity and position of letters (and combinations of letters) from their eye-centered (i.e., retinal) locations. We asked whether location-invariance would emerge from deep unsupervised learning on letter strings and what type of intermediate coding would emerge in the resulting hierarchical generative model. We trained a deep network with three hidden layers on an artificial dataset of letter strings presented at five possible retinal locations. Though word-level information (i.e., word identity) was never provided to the network during training, linear decoding from the activity of the deepest hidden layer yielded near-perfect accuracy in location-invariant word recognition. Conversely, decoding from lower layers yielded a large number of transposition errors. Analyses of emergent internal representations showed that word selectivity and location invariance increased as a function of layer depth. Word-tuning and location-invariance were found at the level of single neurons, but there was no evidence for bigram coding. Finally, the distributed internal representation of words at the deepest layer showed higher similarity to the representation elicited by the two exterior letters than by other combinations of two contiguous letters, in agreement with the hypothesis that word edges have special status. These results reveal that the efficient coding of written words—which was the model's learning objective�

Deep generative learning of location-invariant visual word recognition.

Science.gov (United States)

Di Bono, Maria Grazia; Zorzi, Marco

2013-01-01

It is widely believed that orthographic processing implies an approximate, flexible coding of letter position, as shown by relative-position and transposition priming effects in visual word recognition. These findings have inspired alternative proposals about the representation of letter position, ranging from noisy coding across the ordinal positions to relative position coding based on open bigrams. This debate can be cast within the broader problem of learning location-invariant representations of written words, that is, a coding scheme abstracting the identity and position of letters (and combinations of letters) from their eye-centered (i.e., retinal) locations. We asked whether location-invariance would emerge from deep unsupervised learning on letter strings and what type of intermediate coding would emerge in the resulting hierarchical generative model. We trained a deep network with three hidden layers on an artificial dataset of letter strings presented at five possible retinal locations. Though word-level information (i.e., word identity) was never provided to the network during training, linear decoding from the activity of the deepest hidden layer yielded near-perfect accuracy in location-invariant word recognition. Conversely, decoding from lower layers yielded a large number of transposition errors. Analyses of emergent internal representations showed that word selectivity and location invariance increased as a function of layer depth. Word-tuning and location-invariance were found at the level of single neurons, but there was no evidence for bigram coding. Finally, the distributed internal representation of words at the deepest layer showed higher similarity to the representation elicited by the two exterior letters than by other combinations of two contiguous letters, in agreement with the hypothesis that word edges have special status. These results reveal that the efficient coding of written words-which was the model's learning objective

Deep generative learning of location-invariant visual word recognition

Directory of Open Access Journals (Sweden)

Maria Grazia eDi Bono

2013-09-01

Full Text Available It is widely believed that orthographic processing implies an approximate, flexible coding of letter position, as shown by relative-position and transposition priming effects in visual word recognition. These findings have inspired alternative proposals about the representation of letter position, ranging from noisy coding across the ordinal positions to relative position coding based on open bigrams. This debate can be cast within the broader problem of learning location-invariant representations of written words, that is, a coding scheme abstracting the identity and position of letters (and combinations of letters from their eye-centred (i.e., retinal locations. We asked whether location-invariance would emerge from deep unsupervised learning on letter strings and what type of intermediate coding would emerge in the resulting hierarchical generative model. We trained a deep network with three hidden layers on an artificial dataset of letter strings presented at five possible retinal locations. Though word-level information (i.e., word identity was never provided to the network during training, linear decoding from the activity of the deepest hidden layer yielded near-perfect accuracy in location-invariant word recognition. Conversely, decoding from lower layers yielded a large number of transposition errors. Analyses of emergent internal representations showed that word selectivity and location invariance increased as a function of layer depth. Conversely, there was no evidence for bigram coding. Finally, the distributed internal representation of words at the deepest layer showed higher similarity to the representation elicited by the two exterior letters than by other combinations of two contiguous letters, in agreement with the hypothesis that word edges have special status. These results reveal that the efficient coding of written words – which was the model’s learning objective – is largely based on letter-level information.

Traffic sign recognition based on deep convolutional neural network

Science.gov (United States)

Yin, Shi-hao; Deng, Ji-cai; Zhang, Da-wei; Du, Jing-yuan

2017-11-01

Traffic sign recognition (TSR) is an important component of automated driving systems. It is a rather challenging task to design a high-performance classifier for the TSR system. In this paper, we propose a new method for TSR system based on deep convolutional neural network. In order to enhance the expression of the network, a novel structure (dubbed block-layer below) which combines network-in-network and residual connection is designed. Our network has 10 layers with parameters (block-layer seen as a single layer): the first seven are alternate convolutional layers and block-layers, and the remaining three are fully-connected layers. We train our TSR network on the German traffic sign recognition benchmark (GTSRB) dataset. To reduce overfitting, we perform data augmentation on the training images and employ a regularization method named "dropout". The activation function we employ in our network adopts scaled exponential linear units (SELUs), which can induce self-normalizing properties. To speed up the training, we use an efficient GPU to accelerate the convolutional operation. On the test dataset of GTSRB, we achieve the accuracy rate of 99.67%, exceeding the state-of-the-art results.

Printed wax masks for 254 nm deep-UV pattering of PMMA-based microfluidics

KAUST Repository

Fan, Yiqiang

2012-01-13

This paper reports a new technique for masking deep-UV exposure of poly(methyl methacrylate) (PMMA) using a printed wax mask. This technique provides an inexpensive and bulk fabrication method for PMMA structures. The technique involves the direct printing of the mask onto a polymer sheet using a commercial wax printer. The wax layer was then transferred to a PMMA substrate using a thermal laminator, exposed using deep-UV (with a wavelength of 254 nm), developed in an IPA:water solution, and completed by bonding on a PMMA cap layer. A sample microfluidic device fabricated with this method is also presented, with the microchannel as narrow as 50 μm. The whole process is easy to perform without the requirement for any microfabrication facilities. © 2012 IOP Publishing Ltd.

Carbon storage and nutrient mobilization from soil minerals by deep roots and rhizospheres

DEFF Research Database (Denmark)

Callesen, Ingeborg; Harrison, Robert; Stupak, Inge

2016-01-01

studies on potential release of nutrients due to chemical weathering indicate the importance of root access to deep soil layers. Nutrient release profiles clearly indicate depletion in the top layers and a much higher potential in B and C horizons. Reviewing potential sustainability of nutrient supplies......Roots mobilize nutrients via deep soil penetration and rhizosphere processes inducing weathering of primary minerals. These processes contribute to C transfer to soils and to tree nutrition. Assessments of these characteristics and processes of root systems are important for understanding long......-term supplies of nutrient elements essential for forest growth and resilience. Research and techniques have significantly advanced since Olof Tamm’s 1934 “base mineral index” for Swedish forest soils, and the basic nutrient budget estimates for whole-tree harvesting systems of the 1970s. Recent research...

Resolution of Singularities Introduced by Hierarchical Structure in Deep Neural Networks.

Science.gov (United States)

Nitta, Tohru

2017-10-01

We present a theoretical analysis of singular points of artificial deep neural networks, resulting in providing deep neural network models having no critical points introduced by a hierarchical structure. It is considered that such deep neural network models have good nature for gradient-based optimization. First, we show that there exist a large number of critical points introduced by a hierarchical structure in deep neural networks as straight lines, depending on the number of hidden layers and the number of hidden neurons. Second, we derive a sufficient condition for deep neural networks having no critical points introduced by a hierarchical structure, which can be applied to general deep neural networks. It is also shown that the existence of critical points introduced by a hierarchical structure is determined by the rank and the regularity of weight matrices for a specific class of deep neural networks. Finally, two kinds of implementation methods of the sufficient conditions to have no critical points are provided. One is a learning algorithm that can avoid critical points introduced by the hierarchical structure during learning (called avoidant learning algorithm). The other is a neural network that does not have some critical points introduced by the hierarchical structure as an inherent property (called avoidant neural network).

Effect of deep native defects on ultrasound propagation in TlInS{sub 2} layered crystal

Energy Technology Data Exchange (ETDEWEB)

Seyidov, MirHasan Yu., E-mail: smirhasan@gtu.edu.tr [Department of Physics, Gebze Technical University, 41400 Gebze, Kocaeli (Turkey); Institute of Physics of NAS of Azerbaijan, H. Javid Avenue, 33, AZ-1143 Baku (Azerbaijan); Suleymanov, Rauf A. [Department of Physics, Gebze Technical University, 41400 Gebze, Kocaeli (Turkey); Institute of Physics of NAS of Azerbaijan, H. Javid Avenue, 33, AZ-1143 Baku (Azerbaijan); Odrinsky, Andrei P. [Institute of Technical Acoustics, National Academy of Sciences of Belarus, Lyudnikov Avenue 13, Vitebsk 210717 (Belarus); Kırbaş, Cafer [Department of Physics, Gebze Technical University, 41400 Gebze, Kocaeli (Turkey); The Scientific and Technological Research Council of Turkey, National Metrology Institute (TUBITAK UME), PQ 54 41470 Gebze, Kocaeli (Turkey)

2016-09-15

We have investigated p-type semiconductor–ferroelectric TlInS{sub 2} by means of Photo-Induced Current Transient Spectroscopy (PICTS) technique in the temperature range 77–350 K for the detection of native deep defect levels in TlInS{sub 2}. Five native deep defect levels were detected and their energy levels and capture cross sections were evaluated. Focusing on these data, the influence of these defects on the longitudinal and transverse ultrasound waves propagation as well as the effect of electric field on ultrasound waves were studied at different temperatures. The acoustic properties were investigated by the pulse-echo method. The direct contribution of thermally activated charged defects to the acoustic properties of TlInS{sub 2} was demonstrated. The key role of charged native deep level defects in elastic properties of TlInS{sub 2} was shown.

Adaptive Learning Rule for Hardware-based Deep Neural Networks Using Electronic Synapse Devices

OpenAIRE

Lim, Suhwan; Bae, Jong-Ho; Eum, Jai-Ho; Lee, Sungtae; Kim, Chul-Heung; Kwon, Dongseok; Park, Byung-Gook; Lee, Jong-Ho

2017-01-01

In this paper, we propose a learning rule based on a back-propagation (BP) algorithm that can be applied to a hardware-based deep neural network (HW-DNN) using electronic devices that exhibit discrete and limited conductance characteristics. This adaptive learning rule, which enables forward, backward propagation, as well as weight updates in hardware, is helpful during the implementation of power-efficient and high-speed deep neural networks. In simulations using a three-layer perceptron net...

Analysis of the deep rolling process on turbine blades using the FEM/BEM-coupling

International Nuclear Information System (INIS)

Baecker, V; Klocke, F; Wegner, H; Timmer, A; Grzhibovskis, R; Rjasanow, S

2010-01-01

Highly stressed components of aircraft engines, like turbine blades, have to satisfy stringent requirements regarding durability and reliability. The induction of compressive stresses and strain hardening in their surface layer has proven as a promising method to significantly increase their fatigue resistance. The required surface layer properties can be achieved by deep rolling. The determination of optimal process parameters still requires elaborate experimental set-up and subsequent time- and cost-extensive measurements. In previous works the application of the Finite Element Method (FEM) was proposed as an effective and cost reducing alternative to predict the surface layer state for given process parameters. However, FEM requires very fine mesh in the surface layer to resolve the high stress gradients with sufficient accuracy. The hereby caused high time and memory requirements render an efficient simulation of complete turbine components as impossible. In this article a solution is offered by coupling the FEM with the Boundary Elements Method (BEM). It enables the computing of large scale models at low computational cost and high result accuracy. Different approaches of the FEM/BEM-coupling for the simulation of deep rolling are examined with regard to their stability and required computing time.

Influence of growth temperature and temperature ramps on deep level defect incorporation in m-plane GaN

International Nuclear Information System (INIS)

Armstrong, A. M.; Kelchner, K.; Nakamura, S.; DenBaars, S. P.; Speck, J. S.

2013-01-01

The dependence of deep level defect incorporation in m-plane GaN films grown by metal-organic chemical vapor deposition on bulk m-plane GaN substrates as a function of growth temperature (T g ) and T g ramping method was investigated using deep level optical spectroscopy. Understanding the influence of T g on GaN deep level incorporation is important for InGaN/GaN multi-quantum well (MQW) light emitting diodes (LEDs) and laser diodes (LDs) because GaN quantum barrier (QB) layers are grown much colder than thin film GaN to accommodate InGaN QW growth. Deep level spectra of low T g (800 °C) GaN films grown under QB conditions were compared to deep level spectra of high T g (1150 °C) GaN. Reducing T g , increased the defect density significantly (>50×) through introduction of emergent deep level defects at 2.09 eV and 2.9 eV below the conduction band minimum. However, optimizing growth conditions during the temperature ramp when transitioning from high to low T g substantially reduced the density of these emergent deep levels by approximately 40%. The results suggest that it is important to consider the potential for non-radiative recombination in QBs of LED or LD active regions, and tailoring the transition from high T g GaN growth to active layer growth can mitigate such non-radiative channels

Deep-Sea, Deep-Sequencing: Metabarcoding Extracellular DNA from Sediments of Marine Canyons.

Directory of Open Access Journals (Sweden)

Magdalena Guardiola

Full Text Available Marine sediments are home to one of the richest species pools on Earth, but logistics and a dearth of taxonomic work-force hinders the knowledge of their biodiversity. We characterized α- and β-diversity of deep-sea assemblages from submarine canyons in the western Mediterranean using an environmental DNA metabarcoding. We used a new primer set targeting a short eukaryotic 18S sequence (ca. 110 bp. We applied a protocol designed to obtain extractions enriched in extracellular DNA from replicated sediment corers. With this strategy we captured information from DNA (local or deposited from the water column that persists adsorbed to inorganic particles and buffered short-term spatial and temporal heterogeneity. We analysed replicated samples from 20 localities including 2 deep-sea canyons, 1 shallower canal, and two open slopes (depth range 100-2,250 m. We identified 1,629 MOTUs, among which the dominant groups were Metazoa (with representatives of 19 phyla, Alveolata, Stramenopiles, and Rhizaria. There was a marked small-scale heterogeneity as shown by differences in replicates within corers and within localities. The spatial variability between canyons was significant, as was the depth component in one of the canyons where it was tested. Likewise, the composition of the first layer (1 cm of sediment was significantly different from deeper layers. We found that qualitative (presence-absence and quantitative (relative number of reads data showed consistent trends of differentiation between samples and geographic areas. The subset of exclusively benthic MOTUs showed similar patterns of β-diversity and community structure as the whole dataset. Separate analyses of the main metazoan phyla (in number of MOTUs showed some differences in distribution attributable to different lifestyles. Our results highlight the differentiation that can be found even between geographically close assemblages, and sets the ground for future monitoring and conservation

High-level radioactive waste disposal in the deep ocean

International Nuclear Information System (INIS)

Hill, H.W.

1977-01-01

A joint programme has begun between the Fisheries Laboratory, Lowestoft and the Institute of Oceanographic Sciences, Wormley to study the dispersion of radioactivity in the deep ocean arising from the possible dumping of high level waste on the sea bed in vitrified-glass form which would permit slow leakage over a long term scale. The programme consists firstly of the development of a simple diffusion/advection model for the dispersion of radioactivity in a closed and finite ocean, which overcomes many of the criticisms of the earlier model proposed by Webb and Morley. Preliminary results from this new model are comparable to those of the Webb-Morley model for radio isotopes with half-lives of 10-300 years but are considerably more restrictive outside this range, particularly for those which are much longer-lived. The second part of the programme, towards which the emphasis is directed, concerns the field programme planned to measure the advection and diffusion parameters in the deeper layers of the ocean to provide realistic input parameters to the model and increase our fundamental understanding of the environment in which the radioactive materials may be released. The first cruises of the programme will take place in late 1976 and involve deep current meter deployments and float dispersion experiments around the present NEA dump site with some sediment sampling, so that adsorption experiments can be started on typical deep sea sediments. The programme will expand the number of long-term deep moored stations over the next five years and include further float experiments, CTD profiling, and other physical oceanography. In the second half of the 5-year programme, attempts will be made to measure diffusion parameters in the deeper layers of the ocean using radioactive tracers

How stratospheric are deep stratospheric intrusions? LUAMI 2008

Directory of Open Access Journals (Sweden)

T. Trickl

2016-07-01

Full Text Available A large-scale comparison of water-vapour vertical-sounding instruments took place over central Europe on 17 October 2008, during a rather homogeneous deep stratospheric intrusion event (LUAMI, Lindenberg Upper-Air Methods Intercomparison. The measurements were carried out at four observational sites: Payerne (Switzerland, Bilthoven (the Netherlands, Lindenberg (north-eastern Germany, and the Zugspitze mountain (Garmisch-Partenkichen, German Alps, and by an airborne water-vapour lidar system creating a transect of humidity profiles between all four stations. A high data quality was verified that strongly underlines the scientific findings. The intrusion layer was very dry with a minimum mixing ratios of 0 to 35 ppm on its lower west side, but did not drop below 120 ppm on the higher-lying east side (Lindenberg. The dryness hardens the findings of a preceding study (“Part 1”, Trickl et al., 2014 that, e.g., 73 % of deep intrusions reaching the German Alps and travelling 6 days or less exhibit minimum mixing ratios of 50 ppm and less. These low values reflect values found in the lowermost stratosphere and indicate very slow mixing with tropospheric air during the downward transport to the lower troposphere. The peak ozone values were around 70 ppb, confirming the idea that intrusion layers depart from the lowermost edge of the stratosphere. The data suggest an increase of ozone from the lower to the higher edge of the intrusion layer. This behaviour is also confirmed by stratospheric aerosol caught in the layer. Both observations are in agreement with the idea that sections of the vertical distributions of these constituents in the source region were transferred to central Europe without major change. LAGRANTO trajectory calculations demonstrated a rather shallow outflow from the stratosphere just above the dynamical tropopause, for the first time confirming the conclusions in “Part 1” from the Zugspitze CO observations. The

Modeling language and cognition with deep unsupervised learning: a tutorial overview.

Science.gov (United States)

Zorzi, Marco; Testolin, Alberto; Stoianov, Ivilin P

2013-01-01

Deep unsupervised learning in stochastic recurrent neural networks with many layers of hidden units is a recent breakthrough in neural computation research. These networks build a hierarchy of progressively more complex distributed representations of the sensory data by fitting a hierarchical generative model. In this article we discuss the theoretical foundations of this approach and we review key issues related to training, testing and analysis of deep networks for modeling language and cognitive processing. The classic letter and word perception problem of McClelland and Rumelhart (1981) is used as a tutorial example to illustrate how structured and abstract representations may emerge from deep generative learning. We argue that the focus on deep architectures and generative (rather than discriminative) learning represents a crucial step forward for the connectionist modeling enterprise, because it offers a more plausible model of cortical learning as well as a way to bridge the gap between emergentist connectionist models and structured Bayesian models of cognition.

Modeling Language and Cognition with Deep Unsupervised Learning:A Tutorial Overview

Directory of Open Access Journals (Sweden)

Marco eZorzi

2013-08-01

Full Text Available Deep unsupervised learning in stochastic recurrent neural networks with many layers of hidden units is a recent breakthrough in neural computation research. These networks build a hierarchy of progressively more complex distributed representations of the sensory data by fitting a hierarchical generative model. In this article we discuss the theoretical foundations of this approach and we review key issues related to training, testing and analysis of deep networks for modeling language and cognitive processing. The classic letter and word perception problem of McClelland and Rumelhart (1981 is used as a tutorial example to illustrate how structured and abstract representations may emerge from deep generative learning. We argue that the focus on deep architectures and generative (rather than discriminative learning represents a crucial step forward for the connectionist modeling enterprise, because it offers a more plausible model of cortical learning as well as way to bridge the gap between emergentist connectionist models and structured Bayesian models of cognition.

Modeling language and cognition with deep unsupervised learning: a tutorial overview

Science.gov (United States)

Zorzi, Marco; Testolin, Alberto; Stoianov, Ivilin P.

2013-01-01

Deep unsupervised learning in stochastic recurrent neural networks with many layers of hidden units is a recent breakthrough in neural computation research. These networks build a hierarchy of progressively more complex distributed representations of the sensory data by fitting a hierarchical generative model. In this article we discuss the theoretical foundations of this approach and we review key issues related to training, testing and analysis of deep networks for modeling language and cognitive processing. The classic letter and word perception problem of McClelland and Rumelhart (1981) is used as a tutorial example to illustrate how structured and abstract representations may emerge from deep generative learning. We argue that the focus on deep architectures and generative (rather than discriminative) learning represents a crucial step forward for the connectionist modeling enterprise, because it offers a more plausible model of cortical learning as well as a way to bridge the gap between emergentist connectionist models and structured Bayesian models of cognition. PMID:23970869

Active constrained layer damping treatments for shell structures: a deep-shell theory, some intuitive results, and an energy analysis

Science.gov (United States)

Shen, I. Y.

1997-02-01

This paper studies vibration control of a shell structure through use of an active constrained layer (ACL) damping treatment. A deep-shell theory that assumes arbitrary Lamé parameters 0964-1726/6/1/011/img1 and 0964-1726/6/1/011/img2 is first developed. Application of Hamilton's principle leads to the governing Love equations, the charge equation of electrostatics, and the associated boundary conditions. The Love equations and boundary conditions imply that the control action of the ACL for shell treatments consists of two components: free-end boundary actuation and membrane actuation. The free-end boundary actuation is identical to that of beam and plate ACL treatments, while the membrane actuation is unique to shell treatments as a result of the curvatures of the shells. In particular, the membrane actuation may reinforce or counteract the boundary actuation, depending on the location of the ACL treatment. Finally, an energy analysis is developed to determine the proper control law that guarantees the stability of ACL shell treatments. Moreover, the energy analysis results in a simple rule predicting whether or not the membrane actuation reinforces the boundary actuation.

Investigation of epitaxial silicon layers as a material for radiation hardened silicon detectors

International Nuclear Information System (INIS)

Li, Z.; Eremin, V.; Ilyashenko, I.; Ivanov, A.; Verbitskaya, E.

1997-12-01

Epitaxial grown thick layers (≥ 100 micrometers) of high resistivity silicon (Epi-Si) have been investigated as a possible candidate of radiation hardened material for detectors for high-energy physics. As grown Epi-Si layers contain high concentration (up to 2 x 10 12 cm -3 ) of deep levels compared with that in standard high resistivity bulk Si. After irradiation of test diodes by protons (E p = 24 GeV) with a fluence of 1.5 x 10 11 cm -2 , no additional radiation induced deep traps have been detected. A reasonable explanation is that there is a sink of primary radiation induced defects (interstitial and vacancies), possibly by as-grown defects, in epitaxial layers. The ''sinking'' process, however, becomes non-effective at high radiation fluences (10 14 cm -2 ) due to saturation of epitaxial defects by high concentration of radiation induced ones. As a result, at neutron fluence of 1 x 10 14 cm -2 the deep level spectrum corresponds to well-known spectrum of radiation induced defects in high resistivity bulk Si. The net effective concentration in the space charge region equals to 3 x 10 12 cm -3 after 3 months of room temperature storage and reveals similar annealing behavior for epitaxial as compared to bulk silicon

Deep-well injection of liquid radioactive waste in Russia. Present situation

International Nuclear Information System (INIS)

Rybalchenko, A.

1998-01-01

At present there are 3 facilities (polygons) for the deep-well injection of liquid radioactive waste in Russia, all of which were constructed in the mid60's. These facilities are operating successfully, and activities have started in preparation for decommissioning. Liquid radioactive waste is injected into deep porous horizons which act as 'collector-layers', isolated from the surface and from groundwaters by a relatively thick sequence of rock of low permeability. The collector-layers (also collector-horizons) contain salt waters or fresh waters of no practical application, lying beneath the main horizons containing potable waters. Construction of facilities for the deep-well injection of liquid radioactive waste was preceded by geological surveys and investigations which were able to substantiate the feasibility and safety of radioactive waste injection, and to obtain initial data for facility design. Operation of the facilities was accompanied by monitoring which confirmed that the main safety requirement was satisfied i.e. localisation of radioactive waste within specified boundaries of the geologic medium. The opinion of most specialists in the atomic power industry in Russia favours deep-well injection as a solution to the problem of liquid radioactive waste management; during the period of active operation of defence facilities (atomic power industry of the former U.S.S.R.), this disposal method prevented the impact of radioactive waste on man and the environment. The experience accumulated concerning the injection of liquid radioactive waste in Russia is of interest to scientists and engineers engaged in problems of protection and remediation of the environment in the vicinity of nuclear industry facilities; an example of the utilisation of the deep subsurface for solidified radioactive waste and the disposal of different types of nuclear materials. Information on the scientific principles and background for the development of facilities for the injection

On the renewal of the denitrifying layer in the Arabian Sea

Digital Repository Service at National Institute of Oceanography (India)

Somasundar, K.; Naqvi, S.W.A.

A one-dimensional (vertical) advection-diffusion model has been applied to the deep layer characterized by a linear potential temperature(theta)-salinity relationship in the Arabian Sea to estimate the velocity of ascending motion. The results...

Fault Diagnosis Based on Chemical Sensor Data with an Active Deep Neural Network

Science.gov (United States)

Jiang, Peng; Hu, Zhixin; Liu, Jun; Yu, Shanen; Wu, Feng

2016-01-01

Big sensor data provide significant potential for chemical fault diagnosis, which involves the baseline values of security, stability and reliability in chemical processes. A deep neural network (DNN) with novel active learning for inducing chemical fault diagnosis is presented in this study. It is a method using large amount of chemical sensor data, which is a combination of deep learning and active learning criterion to target the difficulty of consecutive fault diagnosis. DNN with deep architectures, instead of shallow ones, could be developed through deep learning to learn a suitable feature representation from raw sensor data in an unsupervised manner using stacked denoising auto-encoder (SDAE) and work through a layer-by-layer successive learning process. The features are added to the top Softmax regression layer to construct the discriminative fault characteristics for diagnosis in a supervised manner. Considering the expensive and time consuming labeling of sensor data in chemical applications, in contrast to the available methods, we employ a novel active learning criterion for the particularity of chemical processes, which is a combination of Best vs. Second Best criterion (BvSB) and a Lowest False Positive criterion (LFP), for further fine-tuning of diagnosis model in an active manner rather than passive manner. That is, we allow models to rank the most informative sensor data to be labeled for updating the DNN parameters during the interaction phase. The effectiveness of the proposed method is validated in two well-known industrial datasets. Results indicate that the proposed method can obtain superior diagnosis accuracy and provide significant performance improvement in accuracy and false positive rate with less labeled chemical sensor data by further active learning compared with existing methods. PMID:27754386

Fault Diagnosis Based on Chemical Sensor Data with an Active Deep Neural Network

Directory of Open Access Journals (Sweden)

Peng Jiang

2016-10-01

Full Text Available Big sensor data provide significant potential for chemical fault diagnosis, which involves the baseline values of security, stability and reliability in chemical processes. A deep neural network (DNN with novel active learning for inducing chemical fault diagnosis is presented in this study. It is a method using large amount of chemical sensor data, which is a combination of deep learning and active learning criterion to target the difficulty of consecutive fault diagnosis. DNN with deep architectures, instead of shallow ones, could be developed through deep learning to learn a suitable feature representation from raw sensor data in an unsupervised manner using stacked denoising auto-encoder (SDAE and work through a layer-by-layer successive learning process. The features are added to the top Softmax regression layer to construct the discriminative fault characteristics for diagnosis in a supervised manner. Considering the expensive and time consuming labeling of sensor data in chemical applications, in contrast to the available methods, we employ a novel active learning criterion for the particularity of chemical processes, which is a combination of Best vs. Second Best criterion (BvSB and a Lowest False Positive criterion (LFP, for further fine-tuning of diagnosis model in an active manner rather than passive manner. That is, we allow models to rank the most informative sensor data to be labeled for updating the DNN parameters during the interaction phase. The effectiveness of the proposed method is validated in two well-known industrial datasets. Results indicate that the proposed method can obtain superior diagnosis accuracy and provide significant performance improvement in accuracy and false positive rate with less labeled chemical sensor data by further active learning compared with existing methods.

Biogeochemical characteristics of suspended particulate matter in deep chlorophyll maximum layers in the southern East China Sea

Directory of Open Access Journals (Sweden)

Q. Liu

2018-04-01

Full Text Available Continental shelves and marginal seas are key sites of particulate organic matter (POM production, remineralization and sequestration, playing an important role in the global carbon cycle. Elemental and stable isotopic compositions of organic carbon and nitrogen are thus frequently used to characterize and distinguish POM and its sources in suspended particles and surface sediments in the marginal seas. Here we investigated suspended particulate matter (SPM collected around deep chlorophyll maximum (DCM layers in the southern East China Sea for particulate organic carbon and nitrogen (POC and PN contents and their isotopic compositions (δ13CPOC and δ15NPN to understand provenance and dynamics of POM. Hydrographic parameters (temperature, salinity and turbidity indicated that the study area was weakly influenced by freshwater derived from the Yangtze River during summer 2013. Elemental and isotopic results showed a large variation in δ13CPOC (−25.8 to −18.2 ‰ and δ15NPN (3.8 to 8.0 ‰, but a narrow molar C ∕ N ratio (4.1–6.3 and low POC ∕ Chl a ratio ( <  200 g g−1 in POM, and indicated that the POM in DCM layers was newly produced by phytoplankton. In addition to temperature effects, the range and distribution of δ13CPOC were controlled by variations in primary productivity and phytoplankton species composition; the former explained  ∼  70 % of the variability in δ13CPOC. However, the variation in δ15NPN was controlled by the nutrient status and δ15NNO3− in seawater, as indicated by similar spatial distribution between δ15NPN and the current pattern and water masses in the East China Sea; although interpretations of δ15NPN data should be verified with the nutrient data in future studies. Furthermore, the POM investigated was weakly influenced by the terrestrial OM supplied by the Yangtze River during summer 2013 due to the reduced sediment supply by the Yangtze River and north

Brine/Rock Interaction in Deep Oceanic Layered Gabbros: Petrological Evidence from Cl-Rich Amphibole, High-Temperature Hydrothermal Veins, and Experiments

Science.gov (United States)

Currin Sala, A. M.; Koepke, J.; Almeev, R. R.; Teagle, D. A. H.; Zihlmann, B.; Wolff, P. E.

2017-12-01

Evidence of high temperature brine/rock interaction is found in hydrothermal veins and dykelets that cross-cut layered olivine gabbros in the deep palaeocrust of the Sumail Ophiolite, Sultanate of Oman. Here we present petrological and geochemical data from these samples, and an experimental attempt to simulate brine/gabbro interaction using externally heated cold seal pressure vessels. The studied natural veins and dykelets contain pargasite, hornblende, actinolite, and Cl-rich pargasite with up to 5 wt% Cl, showing a range of formation conditions from magmatic to metamorphic (hydrothermal) and thus a complex history of brine/rock interaction. In addition, the isotopic study of the radiogenic 87/86Sr and stable 18O in different amphibole types provide an estimate for the extent of seawater influence as alteration agent in the veins of the studied samples. Experiments performed at 750 °C and 200 MPa with different starting materials (chlorine-free amphibole, olivine gabbro powder) and 20 wt% NaCl aqueous brine, illustrate the process by which gabbro-hosted amphibole-rich veins evolve at subsolidus temperatures in the presence of a seawater-derived fluid. Our results demonstrate a decrease in olivine, plagioclase and magnetite content in favour of hastingsite, pargasite and magnesiohornblende, a decrease of IVAl and Ti in the starting amphibole, and an increase in Cl in amphibole, up to 0.2 Cl wt%. Our experiments show the change of magmatic pargasite towards more magnesium and silica-rich end members with results comparable to mildly chlorine-rich pargasites and hornblendes found in the natural samples studied. However, the experimental setup also presents limitations in the attainment of very high-chlorine amphibole (up to 5 wt%). Our analytical and experimental results provide further evidence for the existence of a hydrothermal cooling system in the deep oceanic crust.

Probabilistic analysis of soil : Diaphragm wall friction used for value engineering of deep excavation, north/south metro Amsterdam

NARCIS (Netherlands)

Buykx, S.M.; Delfgaauw, S.; Bosch, J.W.

2009-01-01

The excavation of deep building pits often requires a check against failure by uplift of low permeability ground layers below excavation level. Whenever the weight of these soil layers is less than the pore-water pressure underneath, measures to resist buoyancy are to be considered. The measures

The posterior layer of the thoracolumbar fascia. Its function in load transfer from spine to legs.

NARCIS (Netherlands)

Pool-Goudzwaard, A.L.; Vleeming, A; Stoeckart, R.; Wingerden, Jan Paul; Snijders, Chris

1996-01-01

STUDY DESIGN: The superficial and deep lamina of the posterior layer of the thoracolumbar fascia have been studied anatomically and biomechanically. In embalmed human specimens, the posterior layer has been loaded by simulating the action of various muscles. The effect has been studied using raster

Squeeze-SegNet: a new fast deep convolutional neural network for semantic segmentation

Science.gov (United States)

Nanfack, Geraldin; Elhassouny, Azeddine; Oulad Haj Thami, Rachid

2018-04-01

The recent researches in Deep Convolutional Neural Network have focused their attention on improving accuracy that provide significant advances. However, if they were limited to classification tasks, nowadays with contributions from Scientific Communities who are embarking in this field, they have become very useful in higher level tasks such as object detection and pixel-wise semantic segmentation. Thus, brilliant ideas in the field of semantic segmentation with deep learning have completed the state of the art of accuracy, however this architectures become very difficult to apply in embedded systems as is the case for autonomous driving. We present a new Deep fully Convolutional Neural Network for pixel-wise semantic segmentation which we call Squeeze-SegNet. The architecture is based on Encoder-Decoder style. We use a SqueezeNet-like encoder and a decoder formed by our proposed squeeze-decoder module and upsample layer using downsample indices like in SegNet and we add a deconvolution layer to provide final multi-channel feature map. On datasets like Camvid or City-states, our net gets SegNet-level accuracy with less than 10 times fewer parameters than SegNet.

Object recognition using deep convolutional neural networks with complete transfer and partial frozen layers

NARCIS (Netherlands)

Kruithof, M.C.; Bouma, H.; Fischer, N.M.; Schutte, K.

2016-01-01

Object recognition is important to understand the content of video and allow flexible querying in a large number of cameras, especially for security applications. Recent benchmarks show that deep convolutional neural networks are excellent approaches for object recognition. This paper describes an

The Role of Peat Layers on Iron Dynamics in Peatlands

Directory of Open Access Journals (Sweden)

Arifin Fahmi

2010-09-01

Full Text Available The research aimed to study the effect of peat thickness and humification stage of the peat material on Fe solubility at the peatlands with sulfidic material as substratum. The research was conducted at three conditionals of ombrogen peatlands ie ; deep, moderate and shallow peat. Soil samples were collected by using peat borer according to interlayer (the border layer of peat and mineral layer and conditional of soil horizons. The sample point depth were (cm G.s2 : 25, G.s1 : 50, Int.s : 70, M.s1 : 90 and M.s2 : 100 for shallow peat, G.m2 : 47, G.m1 : 100, Int.m : 120 and M.m1 : 135 for moderate peat and G.d3 : 50, G.d2 : 150, G.d1 : 200, Int.d : 220 and M.d1 : 235 for deep peat respectively. The results showed that most of Fe on the tested soils was found in organic forms. The peat layers above the sulfidic material decreased the Fe2+ solubility at peatlands. Fe2+ concentration in peat layer decreased with its increasing distance from sulfidic material. There was any other processes beside complexation and chelation of Fe2+ by humic material and its processes was reduction of Fe3+ and this conditions was reflected in redox potential values (Eh.

Exploring the isopycnal mixing and helium-heat paradoxes in a suite of Earth System Models

OpenAIRE

A. Gnanadesikan; R. Abernathey; M.-A. Pradal

2014-01-01

This paper uses a suite of Earth system models which simulate the distribution of He isotopes and radiocarbon to examine two paradoxes in Earth science, each of which results from an inconsistency between theoretically motivated global energy balances and direct observations. The helium–heat paradox refers to the fact that helium emissions to the deep ocean are far lower than would be expected given the rate of geothermal heating, since both are thought to b...

A global boundary-layer height climatology

Energy Technology Data Exchange (ETDEWEB)

Dop, H. van; Krol, M.; Holtslag, B. [Inst. for Marine and Atmospheric Research Utrecht, IMAU, Utrecht (Netherlands)

1997-10-01

In principle the ABL (atmospheric boundary layer) height can be retrieved from atmospheric global circulation models since they contain algorithms which determine the intensity of the turbulence as a function of height. However, these data are not routinely available, or on a (vertical) resolution which is too crude in view of the application. This justifies the development of a separate algorithm in order to define the ABL. The algorithm should include the generation of turbulence by both shear and buoyancy and should be based on readily available atmospheric parameters. There is obviously a wide application for boundary heights in off-line global and regional chemistry and transport modelling. It is also a much used parameter in air pollution meteorology. In this article we shall present a theory which is based on current insights in ABL dynamics. The theory is applicable over land and sea surfaces in all seasons. The theory is (for various reasons) not valid in mountainous areas. In areas where boundary-layer clouds or deep cumulus convection are present the theory does not apply. However, the same global atmospheric circulation models contain parameterizations for shallow and deep convection from which separate estimates can be obtained for the extent of vertical mixing. (au)

A system of automated processing of deep water hydrological information

Science.gov (United States)

Romantsov, V. A.; Dyubkin, I. A.; Klyukbin, L. N.

1974-01-01

An automated system for primary and scientific analysis of deep water hydrological information is presented. Primary processing of the data in this system is carried out on a drifting station, which also calculates the parameters of vertical stability of the sea layers, as well as their depths and altitudes. Methods of processing the raw data are described.

Image quality assessment using deep convolutional networks

Science.gov (United States)

Li, Yezhou; Ye, Xiang; Li, Yong

2017-12-01

This paper proposes a method of accurately assessing image quality without a reference image by using a deep convolutional neural network. Existing training based methods usually utilize a compact set of linear filters for learning features of images captured by different sensors to assess their quality. These methods may not be able to learn the semantic features that are intimately related with the features used in human subject assessment. Observing this drawback, this work proposes training a deep convolutional neural network (CNN) with labelled images for image quality assessment. The ReLU in the CNN allows non-linear transformations for extracting high-level image features, providing a more reliable assessment of image quality than linear filters. To enable the neural network to take images of any arbitrary size as input, the spatial pyramid pooling (SPP) is introduced connecting the top convolutional layer and the fully-connected layer. In addition, the SPP makes the CNN robust to object deformations to a certain extent. The proposed method taking an image as input carries out an end-to-end learning process, and outputs the quality of the image. It is tested on public datasets. Experimental results show that it outperforms existing methods by a large margin and can accurately assess the image quality on images taken by different sensors of varying sizes.

Deep Learning for Distribution Channels' Management

Directory of Open Access Journals (Sweden)

Sabina-Cristiana NECULA

2017-01-01

Full Text Available This paper presents an experiment of using deep learning models for distribution channel management. We present an approach that combines self-organizing maps with artificial neural network with multiple hidden layers in order to identify the potential sales that might be addressed for channel distribution change/ management. Our study aims to highlight the evolution of techniques from simple features/learners to more complex learners and feature engineering or sampling techniques. This paper will allow researchers to choose best suited techniques and features to prepare their churn prediction models.

A vertically integrated dynamic RAM-cell: Buried bit line memory cell with floating transfer layer

NARCIS (Netherlands)

Mouthaan, A.J.; Vertregt, Maarten

1986-01-01

A charge injection device has been realized in which charge can be injected on to an MOS-capacitor from a buried layer via an isolated transfer layer. The cell is positioned vertically between word and bit line. LOCOS (local oxidation) is used to isolate the cells and (deep) ion implantation to

A Weighted Deep Representation Learning Model for Imbalanced Fault Diagnosis in Cyber-Physical Systems

Science.gov (United States)

Guo, Yang; Lin, Wenfang; Yu, Shuyang; Ji, Yang

2018-01-01

Predictive maintenance plays an important role in modern Cyber-Physical Systems (CPSs) and data-driven methods have been a worthwhile direction for Prognostics Health Management (PHM). However, two main challenges have significant influences on the traditional fault diagnostic models: one is that extracting hand-crafted features from multi-dimensional sensors with internal dependencies depends too much on expertise knowledge; the other is that imbalance pervasively exists among faulty and normal samples. As deep learning models have proved to be good methods for automatic feature extraction, the objective of this paper is to study an optimized deep learning model for imbalanced fault diagnosis for CPSs. Thus, this paper proposes a weighted Long Recurrent Convolutional LSTM model with sampling policy (wLRCL-D) to deal with these challenges. The model consists of 2-layer CNNs, 2-layer inner LSTMs and 2-Layer outer LSTMs, with under-sampling policy and weighted cost-sensitive loss function. Experiments are conducted on PHM 2015 challenge datasets, and the results show that wLRCL-D outperforms other baseline methods. PMID:29621131

A Weighted Deep Representation Learning Model for Imbalanced Fault Diagnosis in Cyber-Physical Systems

Directory of Open Access Journals (Sweden)

Zhenyu Wu

2018-04-01

Full Text Available Predictive maintenance plays an important role in modern Cyber-Physical Systems (CPSs and data-driven methods have been a worthwhile direction for Prognostics Health Management (PHM. However, two main challenges have significant influences on the traditional fault diagnostic models: one is that extracting hand-crafted features from multi-dimensional sensors with internal dependencies depends too much on expertise knowledge; the other is that imbalance pervasively exists among faulty and normal samples. As deep learning models have proved to be good methods for automatic feature extraction, the objective of this paper is to study an optimized deep learning model for imbalanced fault diagnosis for CPSs. Thus, this paper proposes a weighted Long Recurrent Convolutional LSTM model with sampling policy (wLRCL-D to deal with these challenges. The model consists of 2-layer CNNs, 2-layer inner LSTMs and 2-Layer outer LSTMs, with under-sampling policy and weighted cost-sensitive loss function. Experiments are conducted on PHM 2015 challenge datasets, and the results show that wLRCL-D outperforms other baseline methods.

Detection of Pitting in Gears Using a Deep Sparse Autoencoder

Directory of Open Access Journals (Sweden)

Yongzhi Qu

2017-05-01

Full Text Available In this paper; a new method for gear pitting fault detection is presented. The presented method is developed based on a deep sparse autoencoder. The method integrates dictionary learning in sparse coding into a stacked autoencoder network. Sparse coding with dictionary learning is viewed as an adaptive feature extraction method for machinery fault diagnosis. An autoencoder is an unsupervised machine learning technique. A stacked autoencoder network with multiple hidden layers is considered to be a deep learning network. The presented method uses a stacked autoencoder network to perform the dictionary learning in sparse coding and extract features from raw vibration data automatically. These features are then used to perform gear pitting fault detection. The presented method is validated with vibration data collected from gear tests with pitting faults in a gearbox test rig and compared with an existing deep learning-based approach.

Ocean barrier layers' effect on tropical cyclone intensification.

Science.gov (United States)

Balaguru, Karthik; Chang, Ping; Saravanan, R; Leung, L Ruby; Xu, Zhao; Li, Mingkui; Hsieh, Jen-Shan

2012-09-04

Improving a tropical cyclone's forecast and mitigating its destructive potential requires knowledge of various environmental factors that influence the cyclone's path and intensity. Herein, using a combination of observations and model simulations, we systematically demonstrate that tropical cyclone intensification is significantly affected by salinity-induced barrier layers, which are "quasi-permanent" features in the upper tropical oceans. When tropical cyclones pass over regions with barrier layers, the increased stratification and stability within the layer reduce storm-induced vertical mixing and sea surface temperature cooling. This causes an increase in enthalpy flux from the ocean to the atmosphere and, consequently, an intensification of tropical cyclones. On average, the tropical cyclone intensification rate is nearly 50% higher over regions with barrier layers, compared to regions without. Our finding, which underscores the importance of observing not only the upper-ocean thermal structure but also the salinity structure in deep tropical barrier layer regions, may be a key to more skillful predictions of tropical cyclone intensities through improved ocean state estimates and simulations of barrier layer processes. As the hydrological cycle responds to global warming, any associated changes in the barrier layer distribution must be considered in projecting future tropical cyclone activity.

A novel wavelet sequence based on deep bidirectional LSTM network model for ECG signal classification.

Science.gov (United States)

Yildirim, Özal

2018-05-01

Long-short term memory networks (LSTMs), which have recently emerged in sequential data analysis, are the most widely used type of recurrent neural networks (RNNs) architecture. Progress on the topic of deep learning includes successful adaptations of deep versions of these architectures. In this study, a new model for deep bidirectional LSTM network-based wavelet sequences called DBLSTM-WS was proposed for classifying electrocardiogram (ECG) signals. For this purpose, a new wavelet-based layer is implemented to generate ECG signal sequences. The ECG signals were decomposed into frequency sub-bands at different scales in this layer. These sub-bands are used as sequences for the input of LSTM networks. New network models that include unidirectional (ULSTM) and bidirectional (BLSTM) structures are designed for performance comparisons. Experimental studies have been performed for five different types of heartbeats obtained from the MIT-BIH arrhythmia database. These five types are Normal Sinus Rhythm (NSR), Ventricular Premature Contraction (VPC), Paced Beat (PB), Left Bundle Branch Block (LBBB), and Right Bundle Branch Block (RBBB). The results show that the DBLSTM-WS model gives a high recognition performance of 99.39%. It has been observed that the wavelet-based layer proposed in the study significantly improves the recognition performance of conventional networks. This proposed network structure is an important approach that can be applied to similar signal processing problems. Copyright © 2018 Elsevier Ltd. All rights reserved.

Performance of Deep and Shallow Neural Networks, the Universal Approximation Theorem, Activity Cliffs, and QSAR.

Science.gov (United States)

Winkler, David A; Le, Tu C

2017-01-01

Neural networks have generated valuable Quantitative Structure-Activity/Property Relationships (QSAR/QSPR) models for a wide variety of small molecules and materials properties. They have grown in sophistication and many of their initial problems have been overcome by modern mathematical techniques. QSAR studies have almost always used so-called "shallow" neural networks in which there is a single hidden layer between the input and output layers. Recently, a new and potentially paradigm-shifting type of neural network based on Deep Learning has appeared. Deep learning methods have generated impressive improvements in image and voice recognition, and are now being applied to QSAR and QSAR modelling. This paper describes the differences in approach between deep and shallow neural networks, compares their abilities to predict the properties of test sets for 15 large drug data sets (the kaggle set), discusses the results in terms of the Universal Approximation theorem for neural networks, and describes how DNN may ameliorate or remove troublesome "activity cliffs" in QSAR data sets. © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

Silicon germanium mask for deep silicon etching

KAUST Repository

Serry, Mohamed

2014-07-29

Polycrystalline silicon germanium (SiGe) can offer excellent etch selectivity to silicon during cryogenic deep reactive ion etching in an SF.sub.6/O.sub.2 plasma. Etch selectivity of over 800:1 (Si:SiGe) may be achieved at etch temperatures from -80 degrees Celsius to -140 degrees Celsius. High aspect ratio structures with high resolution may be patterned into Si substrates using SiGe as a hard mask layer for construction of microelectromechanical systems (MEMS) devices and semiconductor devices.

Silicon germanium mask for deep silicon etching

KAUST Repository

Serry, Mohamed; Rubin, Andrew; Refaat, Mohamed; Sedky, Sherif; Abdo, Mohammad

2014-01-01

Polycrystalline silicon germanium (SiGe) can offer excellent etch selectivity to silicon during cryogenic deep reactive ion etching in an SF.sub.6/O.sub.2 plasma. Etch selectivity of over 800:1 (Si:SiGe) may be achieved at etch temperatures from -80 degrees Celsius to -140 degrees Celsius. High aspect ratio structures with high resolution may be patterned into Si substrates using SiGe as a hard mask layer for construction of microelectromechanical systems (MEMS) devices and semiconductor devices.

Some aspects of volcanic ash layers in the Central Indian Basin.

Digital Repository Service at National Institute of Oceanography (India)

Sukumaran, N.P.; Banerjee, R.; Borole, D.V.; Gupta, S.M.

Intercalated volcanic ash layers in two deep-sea sediment cores from the Central Indian Basin (CIB) are examined for the possibility of an in situ source of suboceanic volcanism. An in situ source has been predicated based on the bottom...

DeepFix: A Fully Convolutional Neural Network for Predicting Human Eye Fixations.

Science.gov (United States)

Kruthiventi, Srinivas S S; Ayush, Kumar; Babu, R Venkatesh

2017-09-01

Understanding and predicting the human visual attention mechanism is an active area of research in the fields of neuroscience and computer vision. In this paper, we propose DeepFix, a fully convolutional neural network, which models the bottom-up mechanism of visual attention via saliency prediction. Unlike classical works, which characterize the saliency map using various hand-crafted features, our model automatically learns features in a hierarchical fashion and predicts the saliency map in an end-to-end manner. DeepFix is designed to capture semantics at multiple scales while taking global context into account, by using network layers with very large receptive fields. Generally, fully convolutional nets are spatially invariant-this prevents them from modeling location-dependent patterns (e.g., centre-bias). Our network handles this by incorporating a novel location-biased convolutional layer. We evaluate our model on multiple challenging saliency data sets and show that it achieves the state-of-the-art results.

Event Recognition Based on Deep Learning in Chinese Texts.

Directory of Open Access Journals (Sweden)

Yajun Zhang

Full Text Available Event recognition is the most fundamental and critical task in event-based natural language processing systems. Existing event recognition methods based on rules and shallow neural networks have certain limitations. For example, extracting features using methods based on rules is difficult; methods based on shallow neural networks converge too quickly to a local minimum, resulting in low recognition precision. To address these problems, we propose the Chinese emergency event recognition model based on deep learning (CEERM. Firstly, we use a word segmentation system to segment sentences. According to event elements labeled in the CEC 2.0 corpus, we classify words into five categories: trigger words, participants, objects, time and location. Each word is vectorized according to the following six feature layers: part of speech, dependency grammar, length, location, distance between trigger word and core word and trigger word frequency. We obtain deep semantic features of words by training a feature vector set using a deep belief network (DBN, then analyze those features in order to identify trigger words by means of a back propagation neural network. Extensive testing shows that the CEERM achieves excellent recognition performance, with a maximum F-measure value of 85.17%. Moreover, we propose the dynamic-supervised DBN, which adds supervised fine-tuning to a restricted Boltzmann machine layer by monitoring its training performance. Test analysis reveals that the new DBN improves recognition performance and effectively controls the training time. Although the F-measure increases to 88.11%, the training time increases by only 25.35%.

Event Recognition Based on Deep Learning in Chinese Texts.

Science.gov (United States)

Zhang, Yajun; Liu, Zongtian; Zhou, Wen

2016-01-01

Event recognition is the most fundamental and critical task in event-based natural language processing systems. Existing event recognition methods based on rules and shallow neural networks have certain limitations. For example, extracting features using methods based on rules is difficult; methods based on shallow neural networks converge too quickly to a local minimum, resulting in low recognition precision. To address these problems, we propose the Chinese emergency event recognition model based on deep learning (CEERM). Firstly, we use a word segmentation system to segment sentences. According to event elements labeled in the CEC 2.0 corpus, we classify words into five categories: trigger words, participants, objects, time and location. Each word is vectorized according to the following six feature layers: part of speech, dependency grammar, length, location, distance between trigger word and core word and trigger word frequency. We obtain deep semantic features of words by training a feature vector set using a deep belief network (DBN), then analyze those features in order to identify trigger words by means of a back propagation neural network. Extensive testing shows that the CEERM achieves excellent recognition performance, with a maximum F-measure value of 85.17%. Moreover, we propose the dynamic-supervised DBN, which adds supervised fine-tuning to a restricted Boltzmann machine layer by monitoring its training performance. Test analysis reveals that the new DBN improves recognition performance and effectively controls the training time. Although the F-measure increases to 88.11%, the training time increases by only 25.35%.

Removal of contaminated asphalt layers by using heat generating powder metallic systems

International Nuclear Information System (INIS)

Barinov, A.S.; Karlina, O.K.; Ojovan, M.I.

1996-01-01

Heat generating systems on the base of powder metallic fuel were used for the removal of contaminated asphalt layers. Decontamination of spots which had complex geometric form was performed. Asphalt layers with deep contamination were removed essentially all radionuclides being retained in asphalt residue. Only a small part (1 - 2 %) of radionuclides could pass to combustion slag. No radionuclides were detected in aerosol-gas phase during decontamination process

Robust Individual-Cell/Object Tracking via PCANet Deep Network in Biomedicine and Computer Vision

Directory of Open Access Journals (Sweden)

Bineng Zhong

2016-01-01

Full Text Available Tracking individual-cell/object over time is important in understanding drug treatment effects on cancer cells and video surveillance. A fundamental problem of individual-cell/object tracking is to simultaneously address the cell/object appearance variations caused by intrinsic and extrinsic factors. In this paper, inspired by the architecture of deep learning, we propose a robust feature learning method for constructing discriminative appearance models without large-scale pretraining. Specifically, in the initial frames, an unsupervised method is firstly used to learn the abstract feature of a target by exploiting both classic principal component analysis (PCA algorithms with recent deep learning representation architectures. We use learned PCA eigenvectors as filters and develop a novel algorithm to represent a target by composing of a PCA-based filter bank layer, a nonlinear layer, and a patch-based pooling layer, respectively. Then, based on the feature representation, a neural network with one hidden layer is trained in a supervised mode to construct a discriminative appearance model. Finally, to alleviate the tracker drifting problem, a sample update scheme is carefully designed to keep track of the most representative and diverse samples during tracking. We test the proposed tracking method on two standard individual cell/object tracking benchmarks to show our tracker's state-of-the-art performance.

Voxel-based measurement sensitivity of spatially resolved near-infrared spectroscopy in layered tissues.

Science.gov (United States)

Niwayama, Masatsugu

2018-03-01

We quantitatively investigated the measurement sensitivity of spatially resolved spectroscopy (SRS) across six tissue models: cerebral tissue, a small animal brain, the forehead of a fetus, an adult brain, forearm muscle, and thigh muscle. The optical path length in the voxel of the model was analyzed using Monte Carlo simulations. It was found that the measurement sensitivity can be represented as the product of the change in the absorption coefficient and the difference in optical path length in two states with different source-detector distances. The results clarified the sensitivity ratio between the surface layer and the deep layer at each source-detector distance for each model and identified changes in the deep measurement area when one of the detectors was close to the light source. A comparison was made with the results from continuous-wave spectroscopy. The study also identified measurement challenges that arise when the surface layer is inhomogeneous. Findings on the measurement sensitivity of SRS at each voxel and in each layer can support the correct interpretation of measured values when near-infrared oximetry or functional near-infrared spectroscopy is used to investigate different tissue structures. (2018) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE).

Trophic Diversity of Plankton in the Epipelagic and Mesopelagic Layers of the Tropical and Equatorial Atlantic Determined with Stable Isotopes

Directory of Open Access Journals (Sweden)

Antonio Bode

2018-06-01

Full Text Available Plankton living in the deep ocean either migrate to the surface to feed or feed in situ on other organisms and detritus. Planktonic communities in the upper 800 m of the tropical and equatorial Atlantic were studied using the natural abundance of stable carbon and nitrogen isotopes to identify their food sources and trophic diversity. Seston and zooplankton (>200 µm samples were collected with Niskin bottles and MOCNESS nets, respectively, in the epipelagic (0–200 m, upper mesopelagic (200–500 m, and lower mesopelagic layers (500–800 m at 11 stations. Food sources for plankton in the productive zone influenced by the NW African upwelling and the Canary Current were different from those in the oligotrophic tropical and equatorial zones. In the latter, zooplankton collected during the night in the mesopelagic layers was enriched in heavy nitrogen isotopes relative to day samples, supporting the active migration of organisms from deep layers. Isotopic niches showed also zonal differences in size (largest in the north, mean trophic diversity (largest in the tropical zone, food sources, and the number of trophic levels (largest in the equatorial zone. The observed changes in niche size and overlap (up to 71% between the mesopelagic layers but <50% between the epipelagic and upper mesopelagic layers support the prevalence of in situ feeding at deep layers in tropical and equatorial zooplankton.

Mechanisms of hypolimnion erosion in a deep lake (Lago Maggiore, N. Italy

Directory of Open Access Journals (Sweden)

Elisabetta A. CARRARA

2010-02-01

Full Text Available Holo-oligomixis is one of the most important hydrodynamic characteristics of deep lakes in temperate regions, especially those of the Southern Alps. It influences such important lake chemical and biological processes as the oxygenation of deep layers, recycling of nutrients, vertical migration of plankton, and reproduction. Analysis of physico-chemical data from Lago Maggiore over the years 1951 – 2008 has shown that in addition to ever active but relatively inefficient convective mixing, three other mechanisms act to oxygenate this lake’s deep waters in winter. These are conveyor belt currents, cold and well-oxygenated tributary inflows that sink down to depths of equal density, and differential cooling of littoral waters that subsequently slide down the lake flanks. Their common outcome is to cause deep erosion of the hypolimnion. Heat content and thermal stability also are affected and are analyzed here in relation to external driving forces, examining in particular how dynamics may be altered by climate change.

Deep-sea bioluminescence blooms after dense water formation at the ocean surface.

Directory of Open Access Journals (Sweden)

Christian Tamburini

Full Text Available The deep ocean is the largest and least known ecosystem on Earth. It hosts numerous pelagic organisms, most of which are able to emit light. Here we present a unique data set consisting of a 2.5-year long record of light emission by deep-sea pelagic organisms, measured from December 2007 to June 2010 at the ANTARES underwater neutrino telescope in the deep NW Mediterranean Sea, jointly with synchronous hydrological records. This is the longest continuous time-series of deep-sea bioluminescence ever recorded. Our record reveals several weeks long, seasonal bioluminescence blooms with light intensity up to two orders of magnitude higher than background values, which correlate to changes in the properties of deep waters. Such changes are triggered by the winter cooling and evaporation experienced by the upper ocean layer in the Gulf of Lion that leads to the formation and subsequent sinking of dense water through a process known as "open-sea convection". It episodically renews the deep water of the study area and conveys fresh organic matter that fuels the deep ecosystems. Luminous bacteria most likely are the main contributors to the observed deep-sea bioluminescence blooms. Our observations demonstrate a consistent and rapid connection between deep open-sea convection and bathypelagic biological activity, as expressed by bioluminescence. In a setting where dense water formation events are likely to decline under global warming scenarios enhancing ocean stratification, in situ observatories become essential as environmental sentinels for the monitoring and understanding of deep-sea ecosystem shifts.

HD-MTL: Hierarchical Deep Multi-Task Learning for Large-Scale Visual Recognition.

Science.gov (United States)

Fan, Jianping; Zhao, Tianyi; Kuang, Zhenzhong; Zheng, Yu; Zhang, Ji; Yu, Jun; Peng, Jinye

2017-02-09

In this paper, a hierarchical deep multi-task learning (HD-MTL) algorithm is developed to support large-scale visual recognition (e.g., recognizing thousands or even tens of thousands of atomic object classes automatically). First, multiple sets of multi-level deep features are extracted from different layers of deep convolutional neural networks (deep CNNs), and they are used to achieve more effective accomplishment of the coarseto- fine tasks for hierarchical visual recognition. A visual tree is then learned by assigning the visually-similar atomic object classes with similar learning complexities into the same group, which can provide a good environment for determining the interrelated learning tasks automatically. By leveraging the inter-task relatedness (inter-class similarities) to learn more discriminative group-specific deep representations, our deep multi-task learning algorithm can train more discriminative node classifiers for distinguishing the visually-similar atomic object classes effectively. Our hierarchical deep multi-task learning (HD-MTL) algorithm can integrate two discriminative regularization terms to control the inter-level error propagation effectively, and it can provide an end-to-end approach for jointly learning more representative deep CNNs (for image representation) and more discriminative tree classifier (for large-scale visual recognition) and updating them simultaneously. Our incremental deep learning algorithms can effectively adapt both the deep CNNs and the tree classifier to the new training images and the new object classes. Our experimental results have demonstrated that our HD-MTL algorithm can achieve very competitive results on improving the accuracy rates for large-scale visual recognition.

Deep-sea bioluminescence blooms after dense water formation at the ocean surface.

Science.gov (United States)

Tamburini, Christian; Canals, Miquel; Durrieu de Madron, Xavier; Houpert, Loïc; Lefèvre, Dominique; Martini, Séverine; D'Ortenzio, Fabrizio; Robert, Anne; Testor, Pierre; Aguilar, Juan Antonio; Samarai, Imen Al; Albert, Arnaud; André, Michel; Anghinolfi, Marco; Anton, Gisela; Anvar, Shebli; Ardid, Miguel; Jesus, Ana Carolina Assis; Astraatmadja, Tri L; Aubert, Jean-Jacques; Baret, Bruny; Basa, Stéphane; Bertin, Vincent; Biagi, Simone; Bigi, Armando; Bigongiari, Ciro; Bogazzi, Claudio; Bou-Cabo, Manuel; Bouhou, Boutayeb; Bouwhuis, Mieke C; Brunner, Jurgen; Busto, José; Camarena, Francisco; Capone, Antonio; Cârloganu, Christina; Carminati, Giada; Carr, John; Cecchini, Stefano; Charif, Ziad; Charvis, Philippe; Chiarusi, Tommaso; Circella, Marco; Coniglione, Rosa; Costantini, Heide; Coyle, Paschal; Curtil, Christian; Decowski, Patrick; Dekeyser, Ivan; Deschamps, Anne; Donzaud, Corinne; Dornic, Damien; Dorosti, Hasankiadeh Q; Drouhin, Doriane; Eberl, Thomas; Emanuele, Umberto; Ernenwein, Jean-Pierre; Escoffier, Stéphanie; Fermani, Paolo; Ferri, Marcelino; Flaminio, Vincenzo; Folger, Florian; Fritsch, Ulf; Fuda, Jean-Luc; Galatà, Salvatore; Gay, Pascal; Giacomelli, Giorgio; Giordano, Valentina; Gómez-González, Juan-Pablo; Graf, Kay; Guillard, Goulven; Halladjian, Garadeb; Hallewell, Gregory; van Haren, Hans; Hartman, Joris; Heijboer, Aart J; Hello, Yann; Hernández-Rey, Juan Jose; Herold, Bjoern; Hößl, Jurgen; Hsu, Ching-Cheng; de Jong, Marteen; Kadler, Matthias; Kalekin, Oleg; Kappes, Alexander; Katz, Uli; Kavatsyuk, Oksana; Kooijman, Paul; Kopper, Claudio; Kouchner, Antoine; Kreykenbohm, Ingo; Kulikovskiy, Vladimir; Lahmann, Robert; Lamare, Patrick; Larosa, Giuseppina; Lattuada, Dario; Lim, Gordon; Presti, Domenico Lo; Loehner, Herbert; Loucatos, Sotiris; Mangano, Salvatore; Marcelin, Michel; Margiotta, Annarita; Martinez-Mora, Juan Antonio; Meli, Athina; Montaruli, Teresa; Moscoso, Luciano; Motz, Holger; Neff, Max; Nezri, Emma Nuel; Palioselitis, Dimitris; Păvălaş, Gabriela E; Payet, Kevin; Payre, Patrice; Petrovic, Jelena; Piattelli, Paolo; Picot-Clemente, Nicolas; Popa, Vlad; Pradier, Thierry; Presani, Eleonora; Racca, Chantal; Reed, Corey; Riccobene, Giorgio; Richardt, Carsten; Richter, Roland; Rivière, Colas; Roensch, Kathrin; Rostovtsev, Andrei; Ruiz-Rivas, Joaquin; Rujoiu, Marius; Russo, Valerio G; Salesa, Francisco; Sánchez-Losa, Augustin; Sapienza, Piera; Schöck, Friederike; Schuller, Jean-Pierre; Schussler, Fabian; Shanidze, Rezo; Simeone, Francesco; Spies, Andreas; Spurio, Maurizio; Steijger, Jos J M; Stolarczyk, Thierry; Taiuti, Mauro G F; Toscano, Simona; Vallage, Bertrand; Van Elewyck, Véronique; Vannoni, Giulia; Vecchi, Manuela; Vernin, Pascal; Wijnker, Guus; Wilms, Jorn; de Wolf, Els; Yepes, Harold; Zaborov, Dmitry; De Dios Zornoza, Juan; Zúñiga, Juan

2013-01-01

The deep ocean is the largest and least known ecosystem on Earth. It hosts numerous pelagic organisms, most of which are able to emit light. Here we present a unique data set consisting of a 2.5-year long record of light emission by deep-sea pelagic organisms, measured from December 2007 to June 2010 at the ANTARES underwater neutrino telescope in the deep NW Mediterranean Sea, jointly with synchronous hydrological records. This is the longest continuous time-series of deep-sea bioluminescence ever recorded. Our record reveals several weeks long, seasonal bioluminescence blooms with light intensity up to two orders of magnitude higher than background values, which correlate to changes in the properties of deep waters. Such changes are triggered by the winter cooling and evaporation experienced by the upper ocean layer in the Gulf of Lion that leads to the formation and subsequent sinking of dense water through a process known as "open-sea convection". It episodically renews the deep water of the study area and conveys fresh organic matter that fuels the deep ecosystems. Luminous bacteria most likely are the main contributors to the observed deep-sea bioluminescence blooms. Our observations demonstrate a consistent and rapid connection between deep open-sea convection and bathypelagic biological activity, as expressed by bioluminescence. In a setting where dense water formation events are likely to decline under global warming scenarios enhancing ocean stratification, in situ observatories become essential as environmental sentinels for the monitoring and understanding of deep-sea ecosystem shifts.

Multimodel analysis of anisotropic diffusive tracer-gas transport in a deep arid unsaturated zone

Science.gov (United States)

Green, Christopher T.; Walvoord, Michelle Ann; Andraski, Brian J.; Striegl, Robert G.; Stonestrom, David A.

2015-01-01

Gas transport in the unsaturated zone affects contaminant flux and remediation, interpretation of groundwater travel times from atmospheric tracers, and mass budgets of environmentally important gases. Although unsaturated zone transport of gases is commonly treated as dominated by diffusion, the characteristics of transport in deep layered sediments remain uncertain. In this study, we use a multimodel approach to analyze results of a gas-tracer (SF6) test to clarify characteristics of gas transport in deep unsaturated alluvium. Thirty-five separate models with distinct diffusivity structures were calibrated to the tracer-test data and were compared on the basis of Akaike Information Criteria estimates of posterior model probability. Models included analytical and numerical solutions. Analytical models provided estimates of bulk-scale apparent diffusivities at the scale of tens of meters. Numerical models provided information on local-scale diffusivities and feasible lithological features producing the observed tracer breakthrough curves. The combined approaches indicate significant anisotropy of bulk-scale diffusivity, likely associated with high-diffusivity layers. Both approaches indicated that diffusivities in some intervals were greater than expected from standard models relating porosity to diffusivity. High apparent diffusivities and anisotropic diffusivity structures were consistent with previous observations at the study site of rapid lateral transport and limited vertical spreading of gas-phase contaminants. Additional processes such as advective oscillations may be involved. These results indicate that gases in deep, layered unsaturated zone sediments can spread laterally more quickly, and produce higher peak concentrations, than predicted by homogeneous, isotropic diffusion models.

On the AlxGa1-xN/AlyGa1-yN/AlxGa1-xN (x>y) p-electron blocking layer to improve the hole injection for AlGaN based deep ultraviolet light-emitting diodes

Science.gov (United States)

Chu, Chunshuang; Tian, Kangkai; Fang, Mengqian; Zhang, Yonghui; Li, Luping; Bi, Wengang; Zhang, Zi-Hui

2018-01-01

This work proposes the [0001] oriented AlGaN-based deep ultraviolet (DUV) light-emitting diode (LED) possessing a specifically designed p-electron blocking layer (p-EBL) to achieve the high internal quantum efficiency. Both electrons and holes can be efficiently injected into the active region by adopting the Al0.60Ga0.40N/Al0.50Ga0.50N/Al0.60Ga0.40N structured p-EBL, in which a p-Al0.50Ga0.50N layer is embedded into the p-EBL. Moreover, the impact of different thicknesses for the p-Al0.50Ga0.50N insertion layer on the hole and electron injections has also been investigated. Compared with the DUV LED with the bulk p-Al0.60Ga0.40N as the EBL, the proposed LED architectures improve the light output power if the thickness of the p-Al0.50Ga0.50N insertion layer is properly designed.

Opportunities and obstacles for deep learning in biology and medicine

Science.gov (United States)

2018-01-01

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems—patient classification, fundamental biological processes and treatment of patients—and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural network's prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine. PMID:29618526

Opportunities and obstacles for deep learning in biology and medicine.

Science.gov (United States)

Ching, Travers; Himmelstein, Daniel S; Beaulieu-Jones, Brett K; Kalinin, Alexandr A; Do, Brian T; Way, Gregory P; Ferrero, Enrico; Agapow, Paul-Michael; Zietz, Michael; Hoffman, Michael M; Xie, Wei; Rosen, Gail L; Lengerich, Benjamin J; Israeli, Johnny; Lanchantin, Jack; Woloszynek, Stephen; Carpenter, Anne E; Shrikumar, Avanti; Xu, Jinbo; Cofer, Evan M; Lavender, Christopher A; Turaga, Srinivas C; Alexandari, Amr M; Lu, Zhiyong; Harris, David J; DeCaprio, Dave; Qi, Yanjun; Kundaje, Anshul; Peng, Yifan; Wiley, Laura K; Segler, Marwin H S; Boca, Simina M; Swamidass, S Joshua; Huang, Austin; Gitter, Anthony; Greene, Casey S

2018-04-01

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems-patient classification, fundamental biological processes and treatment of patients-and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural network's prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine. © 2018 The Authors.

Task-specific feature extraction and classification of fMRI volumes using a deep neural network initialized with a deep belief network: Evaluation using sensorimotor tasks.

Science.gov (United States)

Jang, Hojin; Plis, Sergey M; Calhoun, Vince D; Lee, Jong-Hwan

2017-01-15

Feedforward deep neural networks (DNNs), artificial neural networks with multiple hidden layers, have recently demonstrated a record-breaking performance in multiple areas of applications in computer vision and speech processing. Following the success, DNNs have been applied to neuroimaging modalities including functional/structural magnetic resonance imaging (MRI) and positron-emission tomography data. However, no study has explicitly applied DNNs to 3D whole-brain fMRI volumes and thereby extracted hidden volumetric representations of fMRI that are discriminative for a task performed as the fMRI volume was acquired. Our study applied fully connected feedforward DNN to fMRI volumes collected in four sensorimotor tasks (i.e., left-hand clenching, right-hand clenching, auditory attention, and visual stimulus) undertaken by 12 healthy participants. Using a leave-one-subject-out cross-validation scheme, a restricted Boltzmann machine-based deep belief network was pretrained and used to initialize weights of the DNN. The pretrained DNN was fine-tuned while systematically controlling weight-sparsity levels across hidden layers. Optimal weight-sparsity levels were determined from a minimum validation error rate of fMRI volume classification. Minimum error rates (mean±standard deviation; %) of 6.9 (±3.8) were obtained from the three-layer DNN with the sparsest condition of weights across the three hidden layers. These error rates were even lower than the error rates from the single-layer network (9.4±4.6) and the two-layer network (7.4±4.1). The estimated DNN weights showed spatial patterns that are remarkably task-specific, particularly in the higher layers. The output values of the third hidden layer represented distinct patterns/codes of the 3D whole-brain fMRI volume and encoded the information of the tasks as evaluated from representational similarity analysis. Our reported findings show the ability of the DNN to classify a single fMRI volume based on the

Deep Echo State Network (DeepESN): A Brief Survey

OpenAIRE

Gallicchio, Claudio; Micheli, Alessio

2017-01-01

The study of deep recurrent neural networks (RNNs) and, in particular, of deep Reservoir Computing (RC) is gaining an increasing research attention in the neural networks community. The recently introduced deep Echo State Network (deepESN) model opened the way to an extremely efficient approach for designing deep neural networks for temporal data. At the same time, the study of deepESNs allowed to shed light on the intrinsic properties of state dynamics developed by hierarchical compositions ...

Theoretical study on the inverse modeling of deep body temperature measurement

International Nuclear Information System (INIS)

Huang, Ming; Chen, Wenxi

2012-01-01

We evaluated the theoretical aspects of monitoring the deep body temperature distribution with the inverse modeling method. A two-dimensional model was built based on anatomical structure to simulate the human abdomen. By integrating biophysical and physiological information, the deep body temperature distribution was estimated from cutaneous surface temperature measurements using an inverse quasilinear method. Simulations were conducted with and without the heat effect of blood perfusion in the muscle and skin layers. The results of the simulations showed consistently that the noise characteristics and arrangement of the temperature sensors were the major factors affecting the accuracy of the inverse solution. With temperature sensors of 0.05 °C systematic error and an optimized 16-sensor arrangement, the inverse method could estimate the deep body temperature distribution with an average absolute error of less than 0.20 °C. The results of this theoretical study suggest that it is possible to reconstruct the deep body temperature distribution with the inverse method and that this approach merits further investigation. (paper)

Deep subgap feature in amorphous indium gallium zinc oxide: Evidence against reduced indium

International Nuclear Information System (INIS)

Sallis, Shawn; Williams, Deborah S.; Quackenbush, Nicholas F.; Senger, Mikell; Woicik, Joseph C.; White, Bruce E.; Piper, Louis F.J.

2015-01-01

Amorphous indium gallium zinc oxide (a-IGZO) is the archetypal transparent amorphous oxide semiconductor. Despite the gains made with a-IGZO over amorphous silicon in the last decade, the presence of deep subgap states in a-IGZO active layers facilitate instabilities in thin film transistor properties under negative bias illumination stress. Several candidates could contribute to the formation of states within the band gap. Here, we present evidence against In + lone pair active electrons as the origin of the deep subgap features. No In + species are observed, only In 0 nano-crystallites under certain oxygen deficient growth conditions. Our results further support under coordinated oxygen as the source of the deep subgap states. (copyright 2014 WILEY-VCH Verlag GmbH and Co. KGaA, Weinheim)

MARS14 deep-penetration calculation for the ISIS target station shielding

International Nuclear Information System (INIS)

Nakao, Noriaki; Nunomiya, Tomoya; Iwase, Hiroshi; Nakamura, Takashi

2004-01-01

The calculation of neutron penetration through a thick shield was performed with a three-dimensional multi-layer technique using the MARS14(02) Monte Carlo code to compare with the experimental shielding data in 1998 at the ISIS spallation neutron source facility of Rutherford Appleton Laboratory. In this calculation, secondary particles from a tantalum target bombarded by 800-MeV protons were transmitted through a bulk shield of approximately 3-m-thick iron and 1-m-thick concrete. To accomplish this deep-penetration calculation, a three-dimensional multi-layer technique and energy cut-off method were used considering a spatial statistical balance. Finally, the energy spectra of neutrons behind the very thick shield could be calculated down to the thermal energy with good statistics, and the calculated results typically agree well within a factor of two with the experimental data over a broad energy range. The 12 C(n,2n) 11 C reaction rates behind the bulk shield were also calculated, which agree with the experimental data typically within 60%. These results are quite impressive in calculation accuracy for deep-penetration problem

Cardiac Arrhythmia Classification by Multi-Layer Perceptron and Convolution Neural Networks

Directory of Open Access Journals (Sweden)

Shalin Savalia

2018-05-01

Full Text Available The electrocardiogram (ECG plays an imperative role in the medical field, as it records heart signal over time and is used to discover numerous cardiovascular diseases. If a documented ECG signal has a certain irregularity in its predefined features, this is called arrhythmia, the types of which include tachycardia, bradycardia, supraventricular arrhythmias, and ventricular, etc. This has encouraged us to do research that consists of distinguishing between several arrhythmias by using deep neural network algorithms such as multi-layer perceptron (MLP and convolution neural network (CNN. The TensorFlow library that was established by Google for deep learning and machine learning is used in python to acquire the algorithms proposed here. The ECG databases accessible at PhysioBank.com and kaggle.com were used for training, testing, and validation of the MLP and CNN algorithms. The proposed algorithm consists of four hidden layers with weights, biases in MLP, and four-layer convolution neural networks which map ECG samples to the different classes of arrhythmia. The accuracy of the algorithm surpasses the performance of the current algorithms that have been developed by other cardiologists in both sensitivity and precision.

Cardiac Arrhythmia Classification by Multi-Layer Perceptron and Convolution Neural Networks.

Science.gov (United States)

Savalia, Shalin; Emamian, Vahid

2018-05-04

The electrocardiogram (ECG) plays an imperative role in the medical field, as it records heart signal over time and is used to discover numerous cardiovascular diseases. If a documented ECG signal has a certain irregularity in its predefined features, this is called arrhythmia, the types of which include tachycardia, bradycardia, supraventricular arrhythmias, and ventricular, etc. This has encouraged us to do research that consists of distinguishing between several arrhythmias by using deep neural network algorithms such as multi-layer perceptron (MLP) and convolution neural network (CNN). The TensorFlow library that was established by Google for deep learning and machine learning is used in python to acquire the algorithms proposed here. The ECG databases accessible at PhysioBank.com and kaggle.com were used for training, testing, and validation of the MLP and CNN algorithms. The proposed algorithm consists of four hidden layers with weights, biases in MLP, and four-layer convolution neural networks which map ECG samples to the different classes of arrhythmia. The accuracy of the algorithm surpasses the performance of the current algorithms that have been developed by other cardiologists in both sensitivity and precision.

Enhanced Higgs boson to τ(+)τ(-) search with deep learning.

Science.gov (United States)

Baldi, P; Sadowski, P; Whiteson, D

2015-03-20

The Higgs boson is thought to provide the interaction that imparts mass to the fundamental fermions, but while measurements at the Large Hadron Collider (LHC) are consistent with this hypothesis, current analysis techniques lack the statistical power to cross the traditional 5σ significance barrier without more data. Deep learning techniques have the potential to increase the statistical power of this analysis by automatically learning complex, high-level data representations. In this work, deep neural networks are used to detect the decay of the Higgs boson to a pair of tau leptons. A Bayesian optimization algorithm is used to tune the network architecture and training algorithm hyperparameters, resulting in a deep network of eight nonlinear processing layers that improves upon the performance of shallow classifiers even without the use of features specifically engineered by physicists for this application. The improvement in discovery significance is equivalent to an increase in the accumulated data set of 25%.

Uptake of hazardous radionuclides within layered chalcogenide for environmental protection

Energy Technology Data Exchange (ETDEWEB)

Sengupta, Pranesh, E-mail: praneshsengupta@gmail.com [Materials Science Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India); Dudwadkar, N.L. [Fuel Reprocessing Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India); Vishwanadh, B. [Materials Science Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India); Pulhani, V. [Health Physics Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India); Rao, Rekha [Solid State Physics Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India); Tripathi, S.C. [Fuel Reprocessing Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India); Dey, G.K. [Materials Science Division, Bhabha Atomic Research Centre, Mumbai 400 085 (India)

2014-02-15

Highlights: • Layered chalcogenide with CdI{sub 2} crystal structure prepared by hydrothermal route. • Exploration of the possibilities for radionuclides’ uptake using layered chalcogenide. • Proposing ‘topotactic ionic substitution’ as major uptake mechanism. -- Abstract: Ensuring environmental protection in and around nuclear facilities is a matter of deep concern. Toward this, layered chalcogenide with CdI{sub 2} crystal structure has been prepared. Structural characterizations of layered chalcogenide suggest ‘topotactic ionic substitution’ as the dominant mechanism behind uptake of different cations within its lattice structure. An equilibration time of 45 min and volume to mass ratio of 30:1 are found to absorb {sup 233}U, {sup 239}Pu, {sup 106}Ru, {sup 85+89}Sr, {sup 137}Cs and {sup 241}Am radionuclides to the maximum extents.

Assumed Probability Density Functions for Shallow and Deep Convection

OpenAIRE

Steven K Krueger; Peter A Bogenschutz; Marat Khairoutdinov

2010-01-01

The assumed joint probability density function (PDF) between vertical velocity and conserved temperature and total water scalars has been suggested to be a relatively computationally inexpensive and unified subgrid-scale (SGS) parameterization for boundary layer clouds and turbulent moments. This paper analyzes the performance of five families of PDFs using large-eddy simulations of deep convection, shallow convection, and a transition from stratocumulus to trade wind cumulus. Three of the PD...

The vertical structure of the Saharan boundary layer: Observations and modelling

Science.gov (United States)

Garcia-Carreras, L.; Parker, D. J.; Marsham, J. H.; Rosenberg, P.; Marenco, F.; Mcquaid, J.

2012-04-01

The vertical structure of the Saharan atmospheric boundary layer (SABL) is investigated with the use of aircraft data from the Fennec observational campaign, and high-resolution large-eddy model (LEM) simulations. The SABL is one of the deepest on Earth, and crucial in controlling the vertical redistribution and long-range transport of dust in the Sahara. The SABL is typically made up of an actively growing convective region driven by high sensible heating at the surface, with a deep, near-neutrally stratified Saharan residual layer (SRL) above it, which is mostly well mixed in humidity and temperature and reaches a height of ~500hPa. These two layers are usually separated by a weak (≤1K) temperature inversion, making the vertical structure very sensitive to the surface fluxes. Large-eddy model (LEM) simulations initialized with radiosonde data from Bordj Bardji Mokhtar (BBM), southern Algeria, are used to improve our understanding of the turbulence structure of the stratification of the SABL, and any mixing or exchanges between the different layers. The model can reproduce the typical SABL structure from observations, and a tracer is used to illustrate the growth of the convective boundary layer into the residual layer above. The heat fluxes show a deep entrainment zone between the convective region and the SRL, potentially enhanced by the combination of a weak lid and a neutral layer above. The horizontal variability in the depth of the convective layer was also significant even with homogeneous surface fluxes. Aircraft observations from a number of flights are used to validate the model results, and to highlight the variability present in a more realistic setting, where conditions are rarely homogeneous in space. Stacked legs were performed to get an estimate of the mean flux profile of the boundary layer, as well as the variations in the vertical structure of the SABL with heterogeneous atmospheric and surface conditions. Regular radiosondes from BBM put

[Adhesion of sealer cements to dentin with and without smear layer].

Science.gov (United States)

Gettleman, B H; Messer, H H; ElDeeb, M E

1991-01-01

The influence of a smear layer on the adhesion of sealer cements to dentin was assessed in recently extracted human anterior teeth. A total of 120 samples was tested, 40 per sealer; 20 each with and without the smear layer. The teeth were split longitudinally, and the internal surfaces were ground flat. One-half of each tooth was left with the smear layer intact, while the other half had the smear removed by washing for 3 min with 17% EDTA followed by 5.25% NaOCI. Evidence of the ability to remove the smear layer was verified by scanning electron microscopy. Using a specially designed jig, the sealer was placed into a 4-mm wide x 4 mm deep well which was then set onto the tooth.

Seismic Evaluation of Hydrocarbon Saturation in Deep-Water Reservoirs

Energy Technology Data Exchange (ETDEWEB)

Michael Batzle

2006-04-30

During this last period of the ''Seismic Evaluation of Hydrocarbon Saturation in Deep-Water Reservoirs'' project (Grant/Cooperative Agreement DE-FC26-02NT15342), we finalized integration of rock physics, well log analysis, seismic processing, and forward modeling techniques. Most of the last quarter was spent combining the results from the principal investigators and come to some final conclusions about the project. Also much of the effort was directed towards technology transfer through the Direct Hydrocarbon Indicators mini-symposium at UH and through publications. As a result we have: (1) Tested a new method to directly invert reservoir properties, water saturation, Sw, and porosity from seismic AVO attributes; (2) Constrained the seismic response based on fluid and rock property correlations; (3) Reprocessed seismic data from Ursa field; (4) Compared thin layer property distributions and averaging on AVO response; (5) Related pressures and sorting effects on porosity and their influence on DHI's; (6) Examined and compared gas saturation effects for deep and shallow reservoirs; (7) Performed forward modeling using geobodies from deepwater outcrops; (8) Documented velocities for deepwater sediments; (9) Continued incorporating outcrop descriptive models in seismic forward models; (10) Held an open DHI symposium to present the final results of the project; (11) Relations between Sw, porosity, and AVO attributes; (12) Models of Complex, Layered Reservoirs; and (14) Technology transfer Several factors can contribute to limit our ability to extract accurate hydrocarbon saturations in deep water environments. Rock and fluid properties are one factor, since, for example, hydrocarbon properties will be considerably different with great depths (high pressure) when compared to shallow properties. Significant over pressure, on the other hand will make the rocks behave as if they were shallower. In addition to the physical properties, the scale and

Release pathways for deep seabed disposal of radioactive wastes

International Nuclear Information System (INIS)

Anderson, D.R.; Bishop, W.P.; Bowen, V.T.; Brannen, J.P.; Caudle, W.N.; Detry, R.J.; Ewart, T.E.; Hayes, D.E.; Heath, G.R.; Hessler, R.R.; Hollister, C.D.; Keil, K.; McGowan, J.A.; Rohde, R.W.; Schimmel, W.P.; Schuster, C.L.; Silva, A.J.; Smyrl, W.H.; Taft, B.A.; Talbert, D.M.

1975-01-01

The concept of disposal of high-level solidified and encapsulated radioactive wastes into the deep sea floor has recently been discussed. Such a scheme has conceptual advantages in that the areas of the mid-plate/mid-gyre regions of the oceans are relatively unproductive biologically and relatively devoid of cataclysmic events, and natural processes there are generally quite slow. Given a lack of singular events, a set of barriers against the dispersion of the radioisotopes may be defined. The inverse of these barriers is the set of mechanisms by which the isotopes are transported from the burial site through the barriers to parts of the ocean of immediate significance to mankind. These include: corrosion of the cask; leaching of the waste material; upward transport of the isotopes with the upward moving pore water (mediated by ion-exchange processes); biological transport through bioturbition in the upper sediment layers and lowest water layer; the slow throughput currents of the deep basins; advection and diffusion through the water column; thermally driven transport through the sediments or the water column; biological transport of incorporated isotopes across the seabed or upward through the water column. In principle, the rates of all these processes are measurable or capable of being estimated. Such estimates are given on the basis of present knowledge of the processes in the deep basins. A methodology is discussed for the analytical treatment of the set of processes to give the amount of the isotopes reaching some part of the environment (e.g., and oceanic regime of immediate significance to man) as a function of time. The authors conclude that disposal in the deep seabed is conceptually attractive because of the stability and predictablity of the environment, but that it is not possible to give a firm estimate of the safety of such a scheme from the current knowledge of the mid-plate/mid-gyre regions. (author)

Predicting DNA Methylation State of CpG Dinucleotide Using Genome Topological Features and Deep Networks.

Science.gov (United States)

Wang, Yiheng; Liu, Tong; Xu, Dong; Shi, Huidong; Zhang, Chaoyang; Mo, Yin-Yuan; Wang, Zheng

2016-01-22

The hypo- or hyper-methylation of the human genome is one of the epigenetic features of leukemia. However, experimental approaches have only determined the methylation state of a small portion of the human genome. We developed deep learning based (stacked denoising autoencoders, or SdAs) software named "DeepMethyl" to predict the methylation state of DNA CpG dinucleotides using features inferred from three-dimensional genome topology (based on Hi-C) and DNA sequence patterns. We used the experimental data from immortalised myelogenous leukemia (K562) and healthy lymphoblastoid (GM12878) cell lines to train the learning models and assess prediction performance. We have tested various SdA architectures with different configurations of hidden layer(s) and amount of pre-training data and compared the performance of deep networks relative to support vector machines (SVMs). Using the methylation states of sequentially neighboring regions as one of the learning features, an SdA achieved a blind test accuracy of 89.7% for GM12878 and 88.6% for K562. When the methylation states of sequentially neighboring regions are unknown, the accuracies are 84.82% for GM12878 and 72.01% for K562. We also analyzed the contribution of genome topological features inferred from Hi-C. DeepMethyl can be accessed at http://dna.cs.usm.edu/deepmethyl/.

A New Technique for Deep in situ Measurements of the Soil Water Retention Behaviour

DEFF Research Database (Denmark)

Rocchi, Irene; Gragnano, Carmine Gerardo; Govoni, Laura

2018-01-01

In situ measurements of soil suction and water content in deep soil layers still represent an experimental challenge. Mostly developed within agriculture related disciplines, field techniques for the identification of soil retention behaviour have been so far employed in the geotechnical context ...

Fe gettering by p+ layer in bifacial Si solar cell fabrication

International Nuclear Information System (INIS)

Terakawa, T.; Wang, D.; Nakashima, H.

2006-01-01

Gettering behaviors of Fe into solar cell grade Si are investigated by deep level transient spectroscopy. The samples contaminated with Fe in the range of the concentration of 1.5x10 12 -2.0x10 14 cm -3 were annealed at 600 deg. C to induce gettering. It is shown that the surface layer gettering behaviors of Fe for the sample without p + layer strongly depend on the Fe contamination level, in which the surface layer gettering is not effective for the sample with low level contamination 13 cm -3 but effective for the sample with middle level contamination of 1-5x10 13 cm -3 . In contrast, the samples with p + layer show effective gettering for low and middle level contaminations. The gettering mechanisms in solar cell grade Si without and with p + layer are discussed in details

An Isopycnic Coordinate Numerical Model of the Agulhas Current with Comparison to Observations

Science.gov (United States)

1990-12-01

anomalous temperature signature at the surface due to the intense heat flux from the sea surface in the eastern South Atlantic (Walker and Mey, 1988...to discern since the sea surface temperature signal is erased very quickly due to the high heat flux in the eastern South Atlantic. Consequently...ITERACE0 .... LAYER I P1.0W PATTERN AND INTERFACE DUP’I AT3100DAYS .KNL AT3120DAYS SOUTHI AFRICAI ..* ... SOUTHIAPRICAI... AGUAS BAN AOUUIAS BAN Figure

Why & When Deep Learning Works: Looking Inside Deep Learnings

OpenAIRE

Ronen, Ronny

2017-01-01

The Intel Collaborative Research Institute for Computational Intelligence (ICRI-CI) has been heavily supporting Machine Learning and Deep Learning research from its foundation in 2012. We have asked six leading ICRI-CI Deep Learning researchers to address the challenge of "Why & When Deep Learning works", with the goal of looking inside Deep Learning, providing insights on how deep networks function, and uncovering key observations on their expressiveness, limitations, and potential. The outp...

Identification of deep levels in GaN associated with dislocations

International Nuclear Information System (INIS)

Soh, C B; Chua, S J; Lim, H F; Chi, D Z; Liu, W; Tripathy, S

2004-01-01

To establish a correlation between dislocations and deep levels in GaN, a deep-level transient spectroscopy study has been carried out on GaN samples grown by metalorganic chemical vapour deposition. In addition to typical undoped and Si-doped GaN samples, high-quality crack-free undoped GaN film grown intentionally on heavily doped cracked Si-doped GaN and cracked AlGaN templates are also chosen for this study. The purpose of growth of such continuous GaN layers on top of the cracked templates is to reduce the screw dislocation density by an order of magnitude. Deep levels in these layers have been characterized and compared with emphasis on their thermal stabilities and capture kinetics. Three electron traps at E c -E T ∼0.10-0.11, 0.24-0.27 and 0.59-0.63 eV are detected common to all the samples while additional levels at E c -E T ∼0.18 and 0.37-0.40 eV are also observed in the Si-doped GaN. The trap levels exhibit considerably different stabilities under rapid thermal annealing. Based on the observations, the trap levels at E c -E T ∼0.18 and 0.24-0.27 eV can be associated with screw dislocations, whereas the level at E c -E T ∼0.59-0.63 eV can be associated with edge dislocations. This is also in agreement with the transmission electron microscopy measurements conducted on the GaN samples

Contrasting effects of strabismic amblyopia on metabolic activity in superficial and deep layers of striate cortex.

Science.gov (United States)

Adams, Daniel L; Economides, John R; Horton, Jonathan C

2015-05-01

To probe the mechanism of visual suppression, we have raised macaques with strabismus by disinserting the medial rectus muscle in each eye at 1 mo of age. Typically, this operation produces a comitant, alternating exotropia with normal acuity in each eye. Here we describe an unusual occurrence: the development of severe amblyopia in one eye of a monkey after induction of exotropia. Shortly after surgery, the animal demonstrated a strong fixation preference for the left eye, with apparent suppression of the right eye. Later, behavioral testing showed inability to track or to saccade to targets with the right eye. With the left eye occluded, the animal demonstrated no visually guided behavior. Optokinetic nystagmus was absent in the right eye. Metabolic activity in striate cortex was assessed by processing the tissue for cytochrome oxidase (CO). Amblyopia caused loss of CO in one eye's rows of patches, presumably those serving the blind eye. Layers 4A and 4B showed columns of reduced CO, in register with pale rows of patches in layer 2/3. Layers 4C, 5, and 6 also showed columns of CO activity, but remarkably, comparison with more superficial layers showed a reversal in contrast. In other words, pale CO staining in layers 2/3, 4A, and 4B was aligned with dark CO staining in layers 4C, 5, and 6. No experimental intervention or deprivation paradigm has been reported previously to produce opposite effects on metabolic activity in layers 2/3, 4A, and 4B vs. layers 4C, 5, and 6 within a given eye's columns. Copyright © 2015 the American Physiological Society.

Sunspot drawings handwritten character recognition method based on deep learning

Science.gov (United States)

Zheng, Sheng; Zeng, Xiangyun; Lin, Ganghua; Zhao, Cui; Feng, Yongli; Tao, Jinping; Zhu, Daoyuan; Xiong, Li

2016-05-01

High accuracy scanned sunspot drawings handwritten characters recognition is an issue of critical importance to analyze sunspots movement and store them in the database. This paper presents a robust deep learning method for scanned sunspot drawings handwritten characters recognition. The convolution neural network (CNN) is one algorithm of deep learning which is truly successful in training of multi-layer network structure. CNN is used to train recognition model of handwritten character images which are extracted from the original sunspot drawings. We demonstrate the advantages of the proposed method on sunspot drawings provided by Chinese Academy Yunnan Observatory and obtain the daily full-disc sunspot numbers and sunspot areas from the sunspot drawings. The experimental results show that the proposed method achieves a high recognition accurate rate.

Deep subgap feature in amorphous indium gallium zinc oxide: Evidence against reduced indium

Energy Technology Data Exchange (ETDEWEB)

Sallis, Shawn; Williams, Deborah S. [Materials Science and Engineering, Binghamton University, Binghamton, New York, 13902 (United States); Quackenbush, Nicholas F.; Senger, Mikell [Department of Physics, Applied Physics and Astronomy, Binghamton University, Binghamton, New York, 13902 (United States); Woicik, Joseph C. [Materials Science and Engineering Laboratory, National Institute of Standards and Technology, Gaithersburg, Maryland, 20899 (United States); White, Bruce E.; Piper, Louis F.J. [Materials Science and Engineering, Binghamton University, Binghamton, New York, 13902 (United States); Department of Physics, Applied Physics and Astronomy, Binghamton University, Binghamton, New York, 13902 (United States)

2015-07-15

Amorphous indium gallium zinc oxide (a-IGZO) is the archetypal transparent amorphous oxide semiconductor. Despite the gains made with a-IGZO over amorphous silicon in the last decade, the presence of deep subgap states in a-IGZO active layers facilitate instabilities in thin film transistor properties under negative bias illumination stress. Several candidates could contribute to the formation of states within the band gap. Here, we present evidence against In{sup +} lone pair active electrons as the origin of the deep subgap features. No In{sup +} species are observed, only In{sup 0} nano-crystallites under certain oxygen deficient growth conditions. Our results further support under coordinated oxygen as the source of the deep subgap states. (copyright 2014 WILEY-VCH Verlag GmbH and Co. KGaA, Weinheim)

Proceedings of a workshop on physical oceanography related to the subseabed disposal of high-level nuclear waste

International Nuclear Information System (INIS)

Marietta, M.G.

1981-04-01

At this workshop a group of expert scientists: (1) assessed the current state of knowledge with regard to the physical oceanographic questions that must be answered generally if high level nuclear waste is to be disposed of on or under the seabed; (2) discussed physical oceanographic science necessarily related to the US Subseabed Disposal Program; (3) recommended necessary research; and (4) identified other ongoing programs with which important liaisons should be made and continued. This report is a collection of workshop presentations, and recommendations, and a synthesis of topical group recommendations into a unified statement of research needs. The US Seabed Disposal Program is described. The goal is to assess the technical, environmental and engineering feasibility of seabed disposal. The environmental studies program will assess possible ecosystem and health effects from radionuclides which may be released due to accidental leakage. Discussion on the following topics are also included: bottom boundary layer; mixing across isopycnal surfaces; circulation modeling; mesoscale dispersion; deep circulation of the Pacific Ocean; vertical transport at edges; instrumentation; chemical oceanography; plutonium distribution in the Pacific; biology report; chemical dumping report; and low-level waste report

A Study on DLC Tool Coating for Deep Drawing and Ironing of Stainless Steel

DEFF Research Database (Denmark)

Üstünyagiz, Esmeray; Hafis Sulaiman, Mohd; Christiansen, Peter

2018-01-01

) to replicate industrial ironing of deep drawn, stainless steel parts. Non-hazardous tribo-systems in form of a double layer Diamond-like coated tool applied under dry condition or with an environmentally friendly lubricant were investigated via emulating industrial process conditions in laboratory tests...

Extended deep level defects in Ge-condensed SiGe-on-Insulator structures fabricated using proton and helium implantations

International Nuclear Information System (INIS)

Kwak, D.W.; Lee, D.W.; Oh, J.S.; Lee, Y.H.; Cho, H.Y.

2012-01-01

SiGe-on-Insulator (SGOI) structures were created using the Ge condensation method, where an oxidation process is performed on the SiGe/Si structure. This method involves rapid thermal chemical vapor deposition and H + /He + ion-implantations. Deep level defects in these structures were investigated using deep level transient spectroscopy (DLTS) by varying the pulse injection time. According to the DLTS measurements, a deep level defect induced during the Ge condensation process was found at 0.28 eV above the valence band with a capture cross section of 2.67 × 10 −17 cm 2 , two extended deep levels were also found at 0.54 eV and 0.42 eV above the valence band with capture cross sections of 3.17 × 10 −14 cm 2 and 0.96 × 10 −15 cm 2 , respectively. In the SGOI samples with ion-implantation, the densities of the newly generated defects as well as the existing defects were decreased effectively. Furthermore, the Coulomb barrier heights of the extended deep level defects were drastically reduced. Thus, we suggest that the Ge condensation method with H + ion implantation could reduce deep level defects generated from the condensation and control the electrical properties of the condensed SiGe layers. - Highlights: ► We have fabricated low-defective SiGe-on-Insulator (SGOI) with implantation method. ► H + and He + -ions are used for ion-implantation method. ► We have investigated the deep level defects of SGOI layers. ► Ge condensation method using H + ion implantation could reduce extended defects. ► They could enhance electrical properties.

Highly efficient and simplified phosphorescence white organic light-emitting diodes based on synthesized deep-blue host and orange emitter

Energy Technology Data Exchange (ETDEWEB)

Koo, Ja Ryong; Lee, Seok Jae; Hyung, Gun Woo; Kim, Bo Young; Lee, Dong Hyung [Department of Information Display, Hongik University, Seoul 121-791 (Korea, Republic of); Kim, Woo Young [Department of Green Energy and Semiconductor Engineering, Hoseo University, Asan 336-795 (Korea, Republic of); Lee, Kum Hee [Department of Chemistry, Sungkyunkwan University, Suwon 440-746 (Korea, Republic of); Yoon, Seung Soo, E-mail: ssyoon@skku.edu [Department of Chemistry, Sungkyunkwan University, Suwon 440-746 (Korea, Republic of); Kim, Young Kwan, E-mail: kimyk@hongik.ac.kr [Department of Information Display, Hongik University, Seoul 121-791 (Korea, Republic of)

2013-10-01

The authors have demonstrated a highly efficient and stable phosphorescent white organic light-emitting diode (WOLED), which has been achieved by doping only one orange phosphorescent emitter, Bis(5-benzoyl-2-(4-fluorophenyl)pyridinato-C,N)iridium(III) acetylacetonate into an appropriate deep blue phosphorescent host, 4,4'-bis(4-(triphenylsilyl)phenyl)-1,1'-binaphthyl as an emitting layer (EML). The WOLED has been achieved by effective confinement of triplet excitons to emit a warm white color. The optimized WOLED, with a simple structure as a hole transporting layer-EML-electron transporting layer, showed a maximum luminous efficiency of 22.38 cd/A, a maximum power efficiency of 12.01 lm/W, a maximum external quantum efficiency of 7.32%, and CIEx,y coordinates of (0.38,0.42) at 500 cd/m{sup 2}, respectively. - Highlights: • Highly efficient phosphorescent white organic light-emitting diode (WOLED) • Single emitting layer consists of synthesized deep blue host and orange emitter • The WOLED with high EL efficiencies due to efficient triplet exciton confinement.

Hemodynamic monitoring in different cortical layers with a single fiber optical system

Science.gov (United States)

Yu, Linhui; Noor, M. Sohail; Kiss, Zelma H. T.; Murari, Kartikeya

2018-02-01

Functional monitoring of highly-localized deep brain structures is of great interest. However, due to light scattering, optical methods have limited depth penetration or can only measure from a large volume. In this research, we demonstrate continuous measurement of hemodynamics in different cortical layers in response to thalamic deep brain stimulation (DBS) using a single fiber optical system. A 200-μm-core-diameter multimode fiber is used to deliver and collect light from tissue. The fiber probe can be stereotaxically implanted into the brain region of interest at any depth to measure the di use reflectance spectra from a tissue volume of 0.02-0.03 mm3 near the fiber tip. Oxygenation is then extracted from the reflectance spectra using an algorithm based on Monte Carlo simulations. Measurements were performed on the surface (cortical layer I) and at 1.5 mm depth (cortical layer VI) of the motor cortex in anesthetized rats with thalamic DBS. Preliminary results revealed the oxygenation changes in response to DBS. Moreover, the baseline as well as the stimulus-evoked change in oxygenation were different at the two depths of cortex.

STDP-based spiking deep convolutional neural networks for object recognition.

Science.gov (United States)

Kheradpisheh, Saeed Reza; Ganjtabesh, Mohammad; Thorpe, Simon J; Masquelier, Timothée

2018-03-01

Previous studies have shown that spike-timing-dependent plasticity (STDP) can be used in spiking neural networks (SNN) to extract visual features of low or intermediate complexity in an unsupervised manner. These studies, however, used relatively shallow architectures, and only one layer was trainable. Another line of research has demonstrated - using rate-based neural networks trained with back-propagation - that having many layers increases the recognition robustness, an approach known as deep learning. We thus designed a deep SNN, comprising several convolutional (trainable with STDP) and pooling layers. We used a temporal coding scheme where the most strongly activated neurons fire first, and less activated neurons fire later or not at all. The network was exposed to natural images. Thanks to STDP, neurons progressively learned features corresponding to prototypical patterns that were both salient and frequent. Only a few tens of examples per category were required and no label was needed. After learning, the complexity of the extracted features increased along the hierarchy, from edge detectors in the first layer to object prototypes in the last layer. Coding was very sparse, with only a few thousands spikes per image, and in some cases the object category could be reasonably well inferred from the activity of a single higher-order neuron. More generally, the activity of a few hundreds of such neurons contained robust category information, as demonstrated using a classifier on Caltech 101, ETH-80, and MNIST databases. We also demonstrate the superiority of STDP over other unsupervised techniques such as random crops (HMAX) or auto-encoders. Taken together, our results suggest that the combination of STDP with latency coding may be a key to understanding the way that the primate visual system learns, its remarkable processing speed and its low energy consumption. These mechanisms are also interesting for artificial vision systems, particularly for hardware

A Global Survey of Deep Underground Facilities; Examples of Geotechnical and Engineering Capabilities, Achievements, Challenges (Mines, Shafts, Tunnels, Boreholes, Sites and Underground Facilities for Nuclear Waste and Physics R&D): A Guide to Interactive Global Map Layers, Table Database, References and Notes

International Nuclear Information System (INIS)

Tynan, Mark C.; Russell, Glenn P.; Perry, Frank V.; Kelley, Richard E.; Champenois, Sean T.

2017-01-01

These associated tables, references, notes, and report present a synthesis of some notable geotechnical and engineering information used to create four interactive layer maps for selected: 1) deep mines and shafts; 2) existing, considered or planned radioactive waste management deep underground studies or disposal facilities 3) deep large diameter boreholes, and 4) physics underground laboratories and facilities from around the world. These data are intended to facilitate user access to basic information and references regarding “deep underground” facilities, history, activities, and plans. In general, the interactive maps and database provide each facility’s approximate site location, geology, and engineered features (e.g.: access, geometry, depth, diameter, year of operations, groundwater, lithology, host unit name and age, basin; operator, management organization, geographic data, nearby cultural features, other). Although the survey is not comprehensive, it is representative of many of the significant existing and historical underground facilities discussed in the literature addressing radioactive waste management and deep mined geologic disposal safety systems. The global survey is intended to support and to inform: 1) interested parties and decision makers; 2) radioactive waste disposal and siting option evaluations, and 3) safety case development applicable to any mined geologic disposal facility as a demonstration of historical and current engineering and geotechnical capabilities available for use in deep underground facility siting, planning, construction, operations and monitoring.

A Global Survey of Deep Underground Facilities; Examples of Geotechnical and Engineering Capabilities, Achievements, Challenges (Mines, Shafts, Tunnels, Boreholes, Sites and Underground Facilities for Nuclear Waste and Physics R&D): A Guide to Interactive Global Map Layers, Table Database, References and Notes

Energy Technology Data Exchange (ETDEWEB)

Tynan, Mark C. [Idaho National Lab. (INL), Idaho Falls, ID (United States); Russell, Glenn P. [Idaho National Lab. (INL), Idaho Falls, ID (United States); Perry, Frank V. [Idaho National Lab. (INL), Idaho Falls, ID (United States); Kelley, Richard E. [Idaho National Lab. (INL), Idaho Falls, ID (United States); Champenois, Sean T. [Idaho National Lab. (INL), Idaho Falls, ID (United States)

2017-06-13

These associated tables, references, notes, and report present a synthesis of some notable geotechnical and engineering information used to create four interactive layer maps for selected: 1) deep mines and shafts; 2) existing, considered or planned radioactive waste management deep underground studies or disposal facilities 3) deep large diameter boreholes, and 4) physics underground laboratories and facilities from around the world. These data are intended to facilitate user access to basic information and references regarding “deep underground” facilities, history, activities, and plans. In general, the interactive maps and database provide each facility’s approximate site location, geology, and engineered features (e.g.: access, geometry, depth, diameter, year of operations, groundwater, lithology, host unit name and age, basin; operator, management organization, geographic data, nearby cultural features, other). Although the survey is not comprehensive, it is representative of many of the significant existing and historical underground facilities discussed in the literature addressing radioactive waste management and deep mined geologic disposal safety systems. The global survey is intended to support and to inform: 1) interested parties and decision makers; 2) radioactive waste disposal and siting option evaluations, and 3) safety case development applicable to any mined geologic disposal facility as a demonstration of historical and current engineering and geotechnical capabilities available for use in deep underground facility siting, planning, construction, operations and monitoring.

Layered virus protection for the operations and administrative messaging system

Science.gov (United States)

Cortez, R. H.

2002-01-01

NASA's Deep Space Network (DSN) is critical in supporting the wide variety of operating and plannedunmanned flight projects. For day-to-day operations it relies on email communication between the three Deep Space Communication Complexes (Canberra, Goldstone, Madrid) and NASA's Jet Propulsion Laboratory. The Operations & Administrative Messaging system, based on the Microsoft Windows NTand Exchange platform, provides the infrastructure that is required for reliable, mission-critical messaging. The reliability of this system, however, is threatened by the proliferation of email viruses that continue to spread at alarming rates. A layered approach to email security has been implemented across the DSN to protect against this threat.

[Effects of deep plowing time during the fallow period on water storage-consumption characteristics and wheat yield in dry-land soil.

Science.gov (United States)

Dang, Jian You; Pei, Xue Xia; Zhang, Ding Yi; Wang, Jiao Ai; Zhang, Jing; Wu, Xue Ping

2016-09-01

Through a three-year field trail, effects of deep plowing time during the fallow period on water storage of 0-200 cm soil before sowing, water consumption of growth period, and growth and development of wheat were investigated. Results demonstrated that soil water storage (SWS) of the fallow period was influenced by deep plowing time, precipitation, and rainfall distribution. With postponing the time of deep plowing in the fallow period, SWS was increased firstly, and then decreased. SWS with deep plowing in early or middle of August was 23.9-45.8 mm more than that with deep plowing in mid-July. It would benefit SWS when more precipitation occurred in the fallow period or more rainfall was distributed in August and September. Deep plowing at a proper time could facilitate SWS, N and P absorption of wheat, and the number of stems before winter and the spike number. The yield of wheat with deep plowing in early or middle August was 3.67%-18.2% higher than that with deep plowing in mid-July, and it was positively correlated with water storage of 0-200 cm soil during the fallow period and SWS of each soil layer during the wheat growth period. However, this correlation coefficient would be weakened by adequate rainfall in spring, the critical growing period for wheat. The time of deep plowing mainly affected the water consumption at soil layer of 60-140 cm during wheat growth. Under current farming conditions of south Shanxi, the increased grain yield of wheat could be achieved by combining the measures of high wheat stubble and wheat straw covering for holding soil water and deep plowing between the Beginning of Autumn (August 6th) and the Limit of Heat (August 21st) for promoting soil water penetration characteristics to improve the number of stems before winter and spike.

Improving the performance of AlGaN-based deep-ultraviolet light-emitting diodes using electron blocking layer with a heart-shaped graded Al composition

Science.gov (United States)

Kwon, M. R.; Park, T. H.; Lee, T. H.; Lee, B. R.; Kim, T. G.

2018-04-01

We propose a design for highly efficient AlGaN-based deep-ultraviolet light-emitting diodes (DUV LEDs) using a heart-shaped graded Al composition electron-blocking layer (EBL). This novel structure reduced downward band bending at the interface between the last quantum barrier and the EBL and flattened the electrostatic field in the interlayer between the barriers of the multi-quantum barrier EBL. Consequently, electron leakage was significantly suppressed and hole injection efficiency was found to have improved. The parameter values of simulation were extracted from the experimental data of the reference DUV LEDs. Using the SimuLED, we compared the electrical and optical properties of three structures with different Al compositions in the active region and the EBL. The internal quantum efficiency of the proposed structure was shown to exceed those of the reference DUV LEDs by a factor of 1.9. Additionally, the output power at 20 mA was found to increase by a factor of 2.1.

A Deep Learning Approach to Digitally Stain Optical Coherence Tomography Images of the Optic Nerve Head.

Science.gov (United States)

Devalla, Sripad Krishna; Chin, Khai Sing; Mari, Jean-Martial; Tun, Tin A; Strouthidis, Nicholas G; Aung, Tin; Thiéry, Alexandre H; Girard, Michaël J A

2018-01-01

To develop a deep learning approach to digitally stain optical coherence tomography (OCT) images of the optic nerve head (ONH). A horizontal B-scan was acquired through the center of the ONH using OCT (Spectralis) for one eye of each of 100 subjects (40 healthy and 60 glaucoma). All images were enhanced using adaptive compensation. A custom deep learning network was then designed and trained with the compensated images to digitally stain (i.e., highlight) six tissue layers of the ONH. The accuracy of our algorithm was assessed (against manual segmentations) using the dice coefficient, sensitivity, specificity, intersection over union (IU), and accuracy. We studied the effect of compensation, number of training images, and performance comparison between glaucoma and healthy subjects. For images it had not yet assessed, our algorithm was able to digitally stain the retinal nerve fiber layer + prelamina, the RPE, all other retinal layers, the choroid, and the peripapillary sclera and lamina cribrosa. For all tissues, the dice coefficient, sensitivity, specificity, IU, and accuracy (mean) were 0.84 ± 0.03, 0.92 ± 0.03, 0.99 ± 0.00, 0.89 ± 0.03, and 0.94 ± 0.02, respectively. Our algorithm performed significantly better when compensated images were used for training (P deep learning algorithm can simultaneously stain the neural and connective tissues of the ONH, offering a framework to automatically measure multiple key structural parameters of the ONH that may be critical to improve glaucoma management.

Deep Learning with Dynamic Spiking Neurons and Fixed Feedback Weights.

Science.gov (United States)

Samadi, Arash; Lillicrap, Timothy P; Tweed, Douglas B

2017-03-01

Recent work in computer science has shown the power of deep learning driven by the backpropagation algorithm in networks of artificial neurons. But real neurons in the brain are different from most of these artificial ones in at least three crucial ways: they emit spikes rather than graded outputs, their inputs and outputs are related dynamically rather than by piecewise-smooth functions, and they have no known way to coordinate arrays of synapses in separate forward and feedback pathways so that they change simultaneously and identically, as they do in backpropagation. Given these differences, it is unlikely that current deep learning algorithms can operate in the brain, but we that show these problems can be solved by two simple devices: learning rules can approximate dynamic input-output relations with piecewise-smooth functions, and a variation on the feedback alignment algorithm can train deep networks without having to coordinate forward and feedback synapses. Our results also show that deep spiking networks learn much better if each neuron computes an intracellular teaching signal that reflects that cell's nonlinearity. With this mechanism, networks of spiking neurons show useful learning in synapses at least nine layers upstream from the output cells and perform well compared to other spiking networks in the literature on the MNIST digit recognition task.

A deep auto-encoder model for gene expression prediction.

Science.gov (United States)

Xie, Rui; Wen, Jia; Quitadamo, Andrew; Cheng, Jianlin; Shi, Xinghua

2017-11-17

Gene expression is a key intermediate level that genotypes lead to a particular trait. Gene expression is affected by various factors including genotypes of genetic variants. With an aim of delineating the genetic impact on gene expression, we build a deep auto-encoder model to assess how good genetic variants will contribute to gene expression changes. This new deep learning model is a regression-based predictive model based on the MultiLayer Perceptron and Stacked Denoising Auto-encoder (MLP-SAE). The model is trained using a stacked denoising auto-encoder for feature selection and a multilayer perceptron framework for backpropagation. We further improve the model by introducing dropout to prevent overfitting and improve performance. To demonstrate the usage of this model, we apply MLP-SAE to a real genomic datasets with genotypes and gene expression profiles measured in yeast. Our results show that the MLP-SAE model with dropout outperforms other models including Lasso, Random Forests and the MLP-SAE model without dropout. Using the MLP-SAE model with dropout, we show that gene expression quantifications predicted by the model solely based on genotypes, align well with true gene expression patterns. We provide a deep auto-encoder model for predicting gene expression from SNP genotypes. This study demonstrates that deep learning is appropriate for tackling another genomic problem, i.e., building predictive models to understand genotypes' contribution to gene expression. With the emerging availability of richer genomic data, we anticipate that deep learning models play a bigger role in modeling and interpreting genomics.

An Efficient Implementation of Deep Convolutional Neural Networks for MRI Segmentation.

Science.gov (United States)

Hoseini, Farnaz; Shahbahrami, Asadollah; Bayat, Peyman

2018-02-27

Image segmentation is one of the most common steps in digital image processing, classifying a digital image into different segments. The main goal of this paper is to segment brain tumors in magnetic resonance images (MRI) using deep learning. Tumors having different shapes, sizes, brightness and textures can appear anywhere in the brain. These complexities are the reasons to choose a high-capacity Deep Convolutional Neural Network (DCNN) containing more than one layer. The proposed DCNN contains two parts: architecture and learning algorithms. The architecture and the learning algorithms are used to design a network model and to optimize parameters for the network training phase, respectively. The architecture contains five convolutional layers, all using 3 × 3 kernels, and one fully connected layer. Due to the advantage of using small kernels with fold, it allows making the effect of larger kernels with smaller number of parameters and fewer computations. Using the Dice Similarity Coefficient metric, we report accuracy results on the BRATS 2016, brain tumor segmentation challenge dataset, for the complete, core, and enhancing regions as 0.90, 0.85, and 0.84 respectively. The learning algorithm includes the task-level parallelism. All the pixels of an MR image are classified using a patch-based approach for segmentation. We attain a good performance and the experimental results show that the proposed DCNN increases the segmentation accuracy compared to previous techniques.

Grain boundary layer behavior in ZnO/Si heterostructure

International Nuclear Information System (INIS)

Liu Bingce; Liu Cihui; Yi Bo

2010-01-01

The grain boundary layer behavior in ZnO/Si heterostucture is investigated. The current-voltage (I-V) curves, deep level transient spectra (DLTS) and capacitance-voltage (C-V) curves are measured. The transport currents of ZnO/Si heterojunction are dominated by grain boundary layer as high densities of interfacial states existed. The interesting phenomenon that the crossing of In I-V curves of ZnO/Si heterojunction at various measurement temperatures and the decrease of its effective barrier height with the decrement of temperature are in contradiction with the ideal heterojunction thermal emission model is observed. The details will be discussed in the following. (semiconductor physics)

NH4 Be2 BO3 F2 and γ-Be2 BO3 F: Overcoming the Layering Habit in KBe2 BO3 F2 for the Next-Generation Deep-Ultraviolet Nonlinear Optical Materials.

Science.gov (United States)

Peng, Guang; Ye, Ning; Lin, Zheshuai; Kang, Lei; Pan, Shilie; Zhang, Min; Lin, Chensheng; Long, Xifa; Luo, Min; Chen, Yu; Tang, Yu-Huan; Xu, Feng; Yan, Tao

2018-05-12

KBe 2 BO 3 F 2 (KBBF) is still the only practically usable crystal that can generate deep-ultraviolet (DUV) coherent light by direct second harmonic generation (SHG). However, applications are hindered by layering, leading to difficulty in the growth of thick crystals and compromised mechanical integrity. Despite efforts, it is still a great challenge to discover new nonlinear optical (NLO) materials that overcome the layering while keeping the DUV SHG available. Now, two new DUV NLO beryllium borates have been successfully designed and synthesized, NH 4 Be 2 BO 3 F 2 (ABBF) and γ-Be 2 BO 3 F (γ-BBF), which not only overcome the layering but also can be used as next-generation DUV NLO materials with the shortest type I phase-matching second-harmonic wavelength down to 173.9 nm and 146 nm, respectively. Significantly, γ-BBF is superior to KBBF in all metrics and would be the most outstanding DUV NLO crystal. © 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

Particle fluxes in the deep Eastern Mediterranean basins: the role of ocean vertical velocities

Directory of Open Access Journals (Sweden)

L. Patara

2009-03-01

Full Text Available This paper analyzes the relationship between deep sedimentary fluxes and ocean current vertical velocities in an offshore area of the Ionian Sea, the deepest basin of the Eastern Mediterranean Sea. Sediment trap data are collected at 500 m and 2800 m depth in two successive moorings covering the period September 1999–May 2001. A tight coupling is observed between the upper and deep traps and the estimated particle sinking rates are more than 200 m day⁻¹. The current vertical velocity field is computed from a 1/16°×1/16° Ocean General Circulation Model simulation and from the wind stress curl. Current vertical velocities are larger and more variable than Ekman vertical velocities, yet the general patterns are alike. Current vertical velocities are generally smaller than 1 m day⁻¹: we therefore exclude a direct effect of downward velocities in determining high sedimentation rates. However we find that upward velocities in the subsurface layers of the water column are positively correlated with deep particle fluxes. We thus hypothesize that upwelling would produce an increase in upper ocean nutrient levels – thus stimulating primary production and grazing – a few weeks before an enhanced vertical flux is found in the sediment traps. High particle sedimentation rates may be attained by means of rapidly sinking fecal pellets produced by gelatinous macro-zooplankton. Other sedimentation mechanisms, such as dust deposition, are also considered in explaining large pulses of deep particle fluxes. The fast sinking rates estimated in this study might be an evidence of the efficiency of the biological pump in sequestering organic carbon from the surface layers of the deep Eastern Mediterranean basins.

SchNet - A deep learning architecture for molecules and materials

Science.gov (United States)

Schütt, K. T.; Sauceda, H. E.; Kindermans, P.-J.; Tkatchenko, A.; Müller, K.-R.

2018-06-01

Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

Slow Manifolds and Multiple Equilibria in Stratocumulus-Capped Boundary Layers

Directory of Open Access Journals (Sweden)

Junya Uchida

2010-12-01

Full Text Available In marine stratocumulus-capped boundary layers under strong inversions, the timescale for thermodynamic adjustment is roughly a day, much shorter than the multiday timescale for inversion height adjustment. Slow-manifold analysis is introduced to exploit this timescale separation when boundary layer air columns experience only slow changes in their boundary conditions. Its essence is that the thermodynamic structure of the boundary layer remains approximately slaved to its inversion height and the instantaneous boundary conditions; this slaved structure determines the entrainment rate and hence the slow evolution of the inversion height. Slow-manifold analysis is shown to apply to mixed-layer model and large-eddy simulations of an idealized nocturnal stratocumulus- capped boundary layer; simulations with different initial inversion heights collapse onto single relationships of cloud properties with inversion height. Depending on the initial inversion height, the simulations evolve toward a shallow thin-cloud boundary layer or a deep, well-mixed thick cloud boundary layer. In the large-eddy simulations, these evolutions occur on two separate slow manifolds (one of which becomes unstable if cloud droplet concentration is reduced. Applications to analysis of stratocumulus observations and to pockets of open cells and ship tracks are proposed.

Training Deep Convolutional Neural Networks with Resistive Cross-Point Devices.

Science.gov (United States)

Gokmen, Tayfun; Onen, Murat; Haensch, Wilfried

2017-01-01

In a previous work we have detailed the requirements for obtaining maximal deep learning performance benefit by implementing fully connected deep neural networks (DNN) in the form of arrays of resistive devices. Here we extend the concept of Resistive Processing Unit (RPU) devices to convolutional neural networks (CNNs). We show how to map the convolutional layers to fully connected RPU arrays such that the parallelism of the hardware can be fully utilized in all three cycles of the backpropagation algorithm. We find that the noise and bound limitations imposed by the analog nature of the computations performed on the arrays significantly affect the training accuracy of the CNNs. Noise and bound management techniques are presented that mitigate these problems without introducing any additional complexity in the analog circuits and that can be addressed by the digital circuits. In addition, we discuss digitally programmable update management and device variability reduction techniques that can be used selectively for some of the layers in a CNN. We show that a combination of all those techniques enables a successful application of the RPU concept for training CNNs. The techniques discussed here are more general and can be applied beyond CNN architectures and therefore enables applicability of the RPU approach to a large class of neural network architectures.

Training Deep Convolutional Neural Networks with Resistive Cross-Point Devices

Science.gov (United States)

Gokmen, Tayfun; Onen, Murat; Haensch, Wilfried

2017-01-01

In a previous work we have detailed the requirements for obtaining maximal deep learning performance benefit by implementing fully connected deep neural networks (DNN) in the form of arrays of resistive devices. Here we extend the concept of Resistive Processing Unit (RPU) devices to convolutional neural networks (CNNs). We show how to map the convolutional layers to fully connected RPU arrays such that the parallelism of the hardware can be fully utilized in all three cycles of the backpropagation algorithm. We find that the noise and bound limitations imposed by the analog nature of the computations performed on the arrays significantly affect the training accuracy of the CNNs. Noise and bound management techniques are presented that mitigate these problems without introducing any additional complexity in the analog circuits and that can be addressed by the digital circuits. In addition, we discuss digitally programmable update management and device variability reduction techniques that can be used selectively for some of the layers in a CNN. We show that a combination of all those techniques enables a successful application of the RPU concept for training CNNs. The techniques discussed here are more general and can be applied beyond CNN architectures and therefore enables applicability of the RPU approach to a large class of neural network architectures. PMID:29066942

The different influence of the residual layer on the development of the summer convective boundary layer in two deserts in northwest China

Science.gov (United States)

Lin, Zhao; Bo, Han; Shihua, Lv; Lijuan, Wen; Xianhong, Meng; Zhaoguo, Li

2018-02-01

The development of the atmospheric boundary layer is closely connected with the exchange of momentum, heat, and mass near the Earth's surface, especially for a convective boundary layer (CBL). Besides being modulated by the buoyancy flux near the Earth's surface, some studies point out that a neutrally stratified residual layer is also crucial for the appearance of a deep CBL. To verify the importance of the residual layer, the CBLs over two deserts in northwest China (Badan Jaran and Taklimakan) were investigated. The summer CBL mean depth over the Taklimakan Desert is shallower than that over the Badan Jaran Desert, even when the sensible heat flux of the former is stronger. Meanwhile, the climatological mean residual layer in the Badan Jaran Desert is much deeper and neutrally stratified in summer. Moreover, we found a significant and negative correlation between the lapse rate of the residual layer and the CBL depth over the Badan Jaran Desert. The different lapse rates of the residual layer in the two regions are partly connected with the advection heating from large-scale atmospheric circulation. The advection heating tends to reduce the temperature difference in the 700 to 500-hPa layer over the Badan Jaran Desert, and it increases the stability in the same atmospheric layer over the Taklimakan Desert. The advection due to climatological mean atmospheric circulation is more effective at modulating the lapse rate of the residual layer than from varied circulation. Also, the interannual variation of planetary boundary layer (PBL) height over two deserts was found to covary with the wave train.

Moisture Vertical Structure, Deep Convective Organization, and Convective Transition in the Amazon

Science.gov (United States)

Schiro, K. A.; Neelin, J. D.

2017-12-01

Constraining precipitation processes in climate models with observations is crucial to accurately simulating current climate and reducing uncertainties in future projections. Results from the Green Ocean Amazon (GOAmazon) field campaign (2014-2015) provide evidence that deep convection is strongly controlled by the availability of moisture in the free troposphere over the Amazon, much like over tropical oceans. Entraining plume buoyancy calculations confirm that CWV is a good proxy for the conditional instability of the environment, yet differences in convective onset as a function of CWV exist over land and ocean, as well as seasonally and diurnally over land. This is largely due to variability in the contribution of lower tropospheric humidity to the total column moisture. Boundary layer moisture shows a strong relationship to the onset during the day, which largely disappears during nighttime. Using S-Band radar, these transition statistics are examined separately for unorganized and mesoscale-organized convection, which exhibit sharp increases in probability of occurrence with increasing moisture throughout the column, particularly in the lower free troposphere. Retrievals of vertical velocity from a radar wind profiler indicate updraft velocity and mass flux increasing with height through the lower troposphere. A deep-inflow mixing scheme motivated by this — corresponding to deep inflow of environmental air into a plume that grows with height — provides a weighting of boundary layer and free tropospheric air that yields buoyancies consistent with the observed onset of deep convection across seasons and times of day, across land and ocean sites, and for all convection types. This provides a substantial improvement relative to more traditional constant mixing assumptions, and a dramatic improvement relative to no mixing. Furthermore, it provides relationships that are as strong or stronger for mesoscale-organized convection as for unorganized convection.

Hourly air pollution concentrations and their important predictors over Houston, Texas using deep neural networks: case study of DISCOVER-AQ time period

Science.gov (United States)

Eslami, E.; Choi, Y.; Roy, A.

2017-12-01

Air quality forecasting carried out by chemical transport models often show significant error. This study uses a deep-learning approach over the Houston-Galveston-Brazoria (HGB) area to overcome this forecasting challenge, for the DISCOVER-AQ period (September 2013). Two approaches, deep neural network (DNN) using a Multi-Layer Perceptron (MLP) and Restricted Boltzmann Machine (RBM) were utilized. The proposed approaches analyzed input data by identifying features abstracted from its previous layer using a stepwise method. The approaches predicted hourly ozone and PM in September 2013 using several predictors of prior three days, including wind fields, temperature, relative humidity, cloud fraction, precipitation along with PM, ozone, and NOx concentrations. Model-measurement comparisons for available monitoring sites reported Indexes of Agreement (IOA) of around 0.95 for both DNN and RBM. A standard artificial neural network (ANN) (IOA=0.90) with similar architecture showed poorer performance than the deep networks, clearly demonstrating the superiority of the deep approaches. Additionally, each network (both deep and standard) performed significantly better than a previous CMAQ study, which showed an IOA of less than 0.80. The most influential input variables were identified using their associated weights, which represented the sensitivity of ozone to input parameters. The results indicate deep learning approaches can achieve more accurate ozone forecasting and identify the important input variables for ozone predictions in metropolitan areas.

Applying deep learning technology to automatically identify metaphase chromosomes using scanning microscopic images: an initial investigation

Science.gov (United States)

Qiu, Yuchen; Lu, Xianglan; Yan, Shiju; Tan, Maxine; Cheng, Samuel; Li, Shibo; Liu, Hong; Zheng, Bin

2016-03-01

Automated high throughput scanning microscopy is a fast developing screening technology used in cytogenetic laboratories for the diagnosis of leukemia or other genetic diseases. However, one of the major challenges of using this new technology is how to efficiently detect the analyzable metaphase chromosomes during the scanning process. The purpose of this investigation is to develop a computer aided detection (CAD) scheme based on deep learning technology, which can identify the metaphase chromosomes with high accuracy. The CAD scheme includes an eight layer neural network. The first six layers compose of an automatic feature extraction module, which has an architecture of three convolution-max-pooling layer pairs. The 1st, 2nd and 3rd pair contains 30, 20, 20 feature maps, respectively. The seventh and eighth layers compose of a multiple layer perception (MLP) based classifier, which is used to identify the analyzable metaphase chromosomes. The performance of new CAD scheme was assessed by receiver operation characteristic (ROC) method. A number of 150 regions of interest (ROIs) were selected to test the performance of our new CAD scheme. Each ROI contains either interphase cell or metaphase chromosomes. The results indicate that new scheme is able to achieve an area under the ROC curve (AUC) of 0.886+/-0.043. This investigation demonstrates that applying a deep learning technique may enable to significantly improve the accuracy of the metaphase chromosome detection using a scanning microscopic imaging technology in the future.

Design of an elastin-layered dermal regeneration template.

Science.gov (United States)

Mithieux, Suzanne M; Weiss, Anthony S

2017-04-01

We demonstrate a novel approach for the production of tunable quantities of elastic fibers. We also show that exogenous tropoelastin is rate-limiting for elastin synthesis regardless of the age of the dermal fibroblast donor. Additionally, we provide a strategy to further enhance synthesis by older cells through the application of conditioned media. We show that this approach delivers an elastin layer on one side of the leading dermal repair template for contact with the deep dermis in order to deliver prefabricated elastic fibers to a physiologically appropriate site during subsequent surgery. This system is attractive because it provides for the first time a viable path for sufficient, histologically detectable levels of patient elastin into full-thickness wound sites that have until now lacked this elastic underlayer. The scars of full thickness wounds typically lack elasticity. Elastin is essential for skin elasticity and is enriched in the deep dermis. This paper is significant because it shows that: (1) we can generate elastic fibers in tunable quantities, (2) tropoelastin is the rate-limiting component in elastin synthesis in vitro, (3) we can generate elastin fibers regardless of donor age, (4) we describe a novel approach to further increase the numbers and thickness of elastic fibers for older donors, (5) we improve on Integra Dermal Regeneration Template and generate a new hybrid biomaterial intended to subsequently surgically deliver these elastic fibers, (6) the elastic fiber layer is presented on the side of Integra that is intended for delivery into its physiologically appropriate site i.e. the deep dermis. Copyright © 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

A 3D Vertically Integrated Deep N-Well CMOS MAPS for the SuperB Layer0

International Nuclear Information System (INIS)

Traversi, G; Manghisoni, M; Re, V; Gaioni, L; Ratti, L

2011-01-01

Deep N-Well (DNW) Monolithic Active Pixel Sensors (MAPS) have been developed in the last few years with the aim of building monolithic sensors with similar functionalities as hybrid pixels systems. In these devices the triple well option, available in deep submicron processes, is exploited to implement analog and digital signal processing at the pixel level. Many prototypes have been fabricated in a planar (2D) 130nm CMOS technology. A new kind of DNW-MAPS, namely Apsel5 3 D, which exploits the capabilities of vertical integration (3D) processes, is presented and discussed in this paper. The impact of 3D processes on the design and performance of DNW pixel sensors could be large, with significant advantages in terms of detection efficiency, pixel cell size and immunity to cross-talk, therefore complying with the severe constraints set by future HEP experiments.

Optimized Structure of the Traffic Flow Forecasting Model With a Deep Learning Approach.

Science.gov (United States)

Yang, Hao-Fan; Dillon, Tharam S; Chen, Yi-Ping Phoebe

2017-10-01

Forecasting accuracy is an important issue for successful intelligent traffic management, especially in the domain of traffic efficiency and congestion reduction. The dawning of the big data era brings opportunities to greatly improve prediction accuracy. In this paper, we propose a novel model, stacked autoencoder Levenberg-Marquardt model, which is a type of deep architecture of neural network approach aiming to improve forecasting accuracy. The proposed model is designed using the Taguchi method to develop an optimized structure and to learn traffic flow features through layer-by-layer feature granulation with a greedy layerwise unsupervised learning algorithm. It is applied to real-world data collected from the M6 freeway in the U.K. and is compared with three existing traffic predictors. To the best of our knowledge, this is the first time that an optimized structure of the traffic flow forecasting model with a deep learning approach is presented. The evaluation results demonstrate that the proposed model with an optimized structure has superior performance in traffic flow forecasting.

Deep iCrawl: An Intelligent Vision-Based Deep Web Crawler

OpenAIRE

R.Anita; V.Ganga Bharani; N.Nityanandam; Pradeep Kumar Sahoo

2011-01-01

The explosive growth of World Wide Web has posed a challenging problem in extracting relevant data. Traditional web crawlers focus only on the surface web while the deep web keeps expanding behind the scene. Deep web pages are created dynamically as a result of queries posed to specific web databases. The structure of the deep web pages makes it impossible for traditional web crawlers to access deep web contents. This paper, Deep iCrawl, gives a novel and vision-based app...

The Sinking and Spreading of The Antarctic Deep Ice Shelf Water In The Ross Sea Studied By In Situ Observaions and Numerical Modeling

Science.gov (United States)

Rubino, A.; Budillon, G.; Pierini, S.; Spezie, G.

The sinking and spreading of the Deep Ice Shelf Water (DISW) in the Ross Sea are analyzed using in situ observations and the results of a nonlinear, reduced-gravity, frontal layered numerical "plume" model which is able to simulate the motion of a bottom-arrested current over realistic topography. The model is forced by prescribing the thickness of the DISW vein as well as its density structure at the southern model boundary. The ambient temperature and salinity are imposed using hydrographic data acquired by the Italian PNRA-CLIMA project. In the model water of the quiescent ambient ocean is allowed to entrain in the active deep layer due to a simple param- eterization of turbulent mixing. The importance of forcing the model with a realistic ambient density is demonstrated by carrying out a numerical simulation in which the bottom active layer is forced using an idealized ambient density. In a more realis- tic simulation the path and the density structure of the DISW vein flowing over the Challenger Basin are obtained and are found to be in good agreement with data. The evolution of the deep current beyond the continental shelf is also simulated. It provides useful information on the water flow and mixing in a region of the Ross Sea where the paucity of experimental data does not allow for a detailed description of the deep ocean dynamics.

Distribution and diel vertical movements of mesopelagic scattering layers in the Red Sea

KAUST Repository

Klevjer, Thor A.

2012-06-13

The mesopelagic zone of the Red Sea represents an extreme environment due to low food concentrations, high temperatures and low oxygen waters. Nevertheless, a 38 kHz echosounder identified at least four distinct scattering layers during the daytime, of which the 2 deepest layers resided entirely within the mesopelagic zone. Two of the acoustic layers were found above a mesopelagic oxygen minimum zone (OMZ), one layer overlapped with the OMZ, and one layer was found below the OMZ. Almost all organisms in the deep layers migrated to the near-surface waters during the night. Backscatter from a 300 kHz lowered Acoustic Doppler Current Profiler indicated a layer of zooplankton within the OMZ. They carried out DVM, yet a portion remained at mesopelagic depths during the night. Our acoustic measurements showed that the bulk of the acoustic backscatter was restricted to waters shallower than 800 m, suggesting that most of the biomass in the Red Sea resides above this depth. 2012 The Author(s).

Distribution and diel vertical movements of mesopelagic scattering layers in the Red Sea

KAUST Repository

Klevjer, Thor A.; Torres, Daniel J.; Kaartvedt, Stein

2012-01-01

The mesopelagic zone of the Red Sea represents an extreme environment due to low food concentrations, high temperatures and low oxygen waters. Nevertheless, a 38 kHz echosounder identified at least four distinct scattering layers during the daytime, of which the 2 deepest layers resided entirely within the mesopelagic zone. Two of the acoustic layers were found above a mesopelagic oxygen minimum zone (OMZ), one layer overlapped with the OMZ, and one layer was found below the OMZ. Almost all organisms in the deep layers migrated to the near-surface waters during the night. Backscatter from a 300 kHz lowered Acoustic Doppler Current Profiler indicated a layer of zooplankton within the OMZ. They carried out DVM, yet a portion remained at mesopelagic depths during the night. Our acoustic measurements showed that the bulk of the acoustic backscatter was restricted to waters shallower than 800 m, suggesting that most of the biomass in the Red Sea resides above this depth. 2012 The Author(s).

A ground-up construction of deep learning

CERN Multimedia

CERN. Geneva

2015-01-01

I propose to give a ground up construction of deep learning as it is in it's modern state. Starting from it's beginnings in the 90's, I plan on showing the relevant (for physics) differences in optimization, construction, activation functions, initialization, and other tricks that have been accrued over the last 20 years. In addition, I plan on showing why deeper, wider basic feedforward architectures can be used. Coupling this with MaxOut layers, modern GPUs, and including both l1 and l2 forms of regularization, we have the current "state of the art" in basic feedforward networks. I plan on discussing pre-training using deep autoencoders and RBMs, and explaining why this has fallen out of favor when you have lots of labeled data. While discussing each of these points, I propose to explain why these particular characteristics are valuable for HEP. Finally, the last topic on basic feedforward networks -- interpretation. I plan on discussing latent representations of important variables (i.e., mass, pT) that ar...

A deep belief network approach using VDRAS data for nowcasting

Science.gov (United States)

Han, Lei; Dai, Jie; Zhang, Wei; Zhang, Changjiang; Feng, Hanlei

2018-04-01

Nowcasting or very short-term forecasting convective storms is still a challenging problem due to the high nonlinearity and insufficient observation of convective weather. As the understanding of the physical mechanism of convective weather is also insufficient, the numerical weather model cannot predict convective storms well. Machine learning approaches provide a potential way to nowcast convective storms using various meteorological data. In this study, a deep belief network (DBN) is proposed to nowcast convective storms using the real-time re-analysis meteorological data. The nowcasting problem is formulated as a classification problem. The 3D meteorological variables are fed directly to the DBN with dimension of input layer 6*6*80. Three hidden layers are used in the DBN and the dimension of output layer is two. A box-moving method is presented to provide the input features containing the temporal and spatial information. The results show that the DNB can generate reasonable prediction results of the movement and growth of convective storms.

Alterations in Location, Magnitude, and Community Composition of Discrete Layers of Phytoplankton in Cold, Deep Waters Near the 1% Isolume of the Laurentian Great Lake Michigan Among Years With Dramatically Different Meteorological Conditions

Science.gov (United States)

Cuhel, R. L.; Aguilar, C.

2016-02-01

Phytoplankton deep populations have dominated both biomass and productivity in deep basins of Lake Michigan for much of the anthropocene. In recent decades, chronically phosphorus-deficient waters have progressed from lower thermocline diatom assemblages in the 2000s to much deeper picocyanobacterial dominance in the late 2000s. Overwhelming establishment of benthic filter-feeding quagga mussels was instrumental in selection for picoplankton in the 2003-2007 time frame, but in 2008 a return to diatom dominance occurred in conjunction with monumental runoff from the Storm of the Century. Picoplankton gradually returned to significance in ensuing years, but suffered after lakewide ice cover and extremely slow spring warming of winters 2013-2015. Extremely calm summer conditions favored the picoplankton, and a decade of 1% light penetration of 50-60m has consistently enabled very deep productivity by several different divisions of algae. An unusual persistent south wind with basin-scale upwelling stimulated a return of fall diatom bloom for the first time in 2015. Repeated expeditions to offshore deep stations (100-150m) with detailed water sampling based on hydrographic observations often include thin peaks of biogenic silica (diatoms, chrysophytes) offset from one or more distinct layers of picocyanobacteria and mixed eucaryotic phytoplankton. In 2014 large, stable populations of the diatom Tabellaria sp. flourished at 50-60m with highly shade-adapted photosynthetic characteristics but assimilation numbers >1. In 2014-2015, picocyanobacterial maxima moved up in the water column and were dissociated from signals in either in vivo fluorescence or transmission. Physical structure, within-year basin physics sequence timing, and now seemingly ammonium availability may each contribute to phytoplankton ecology in this ocean-scale freshwater ecosystem.

Aero Engine Component Fault Diagnosis Using Multi-Hidden-Layer Extreme Learning Machine with Optimized Structure

Directory of Open Access Journals (Sweden)

Shan Pang

2016-01-01

Full Text Available A new aero gas turbine engine gas path component fault diagnosis method based on multi-hidden-layer extreme learning machine with optimized structure (OM-ELM was proposed. OM-ELM employs quantum-behaved particle swarm optimization to automatically obtain the optimal network structure according to both the root mean square error on training data set and the norm of output weights. The proposed method is applied to handwritten recognition data set and a gas turbine engine diagnostic application and is compared with basic ELM, multi-hidden-layer ELM, and two state-of-the-art deep learning algorithms: deep belief network and the stacked denoising autoencoder. Results show that, with optimized network structure, OM-ELM obtains better test accuracy in both applications and is more robust to sensor noise. Meanwhile it controls the model complexity and needs far less hidden nodes than multi-hidden-layer ELM, thus saving computer memory and making it more efficient to implement. All these advantages make our method an effective and reliable tool for engine component fault diagnosis tool.

Multispectral embedding-based deep neural network for three-dimensional human pose recovery

Science.gov (United States)

Yu, Jialin; Sun, Jifeng

2018-01-01

Monocular image-based three-dimensional (3-D) human pose recovery aims to retrieve 3-D poses using the corresponding two-dimensional image features. Therefore, the pose recovery performance highly depends on the image representations. We propose a multispectral embedding-based deep neural network (MSEDNN) to automatically obtain the most discriminative features from multiple deep convolutional neural networks and then embed their penultimate fully connected layers into a low-dimensional manifold. This compact manifold can explore not only the optimum output from multiple deep networks but also the complementary properties of them. Furthermore, the distribution of each hierarchy discriminative manifold is sufficiently smooth so that the training process of our MSEDNN can be effectively implemented only using few labeled data. Our proposed network contains a body joint detector and a human pose regressor that are jointly trained. Extensive experiments conducted on four databases show that our proposed MSEDNN can achieve the best recovery performance compared with the state-of-the-art methods.

Validation of a high-power, time-resolved, near-infrared spectroscopy system for measurement of superficial and deep muscle deoxygenation during exercise.

Science.gov (United States)

Koga, Shunsaku; Barstow, Thomas J; Okushima, Dai; Rossiter, Harry B; Kondo, Narihiko; Ohmae, Etsuko; Poole, David C

2015-06-01

Near-infrared assessment of skeletal muscle is restricted to superficial tissues due to power limitations of spectroscopic systems. We reasoned that understanding of muscle deoxygenation may be improved by simultaneously interrogating deeper tissues. To achieve this, we modified a high-power (∼8 mW), time-resolved, near-infrared spectroscopy system to increase depth penetration. Precision was first validated using a homogenous optical phantom over a range of inter-optode spacings (OS). Coefficients of variation from 10 measurements were minimal (0.5-1.9%) for absorption (μa), reduced scattering, simulated total hemoglobin, and simulated O2 saturation. Second, a dual-layer phantom was constructed to assess depth sensitivity, and the thickness of the superficial layer was varied. With a superficial layer thickness of 1, 2, 3, and 4 cm (μa = 0.149 cm(-1)), the proportional contribution of the deep layer (μa = 0.250 cm(-1)) to total μa was 80.1, 26.9, 3.7, and 0.0%, respectively (at 6-cm OS), validating penetration to ∼3 cm. Implementation of an additional superficial phantom to simulate adipose tissue further reduced depth sensitivity. Finally, superficial and deep muscle spectroscopy was performed in six participants during heavy-intensity cycle exercise. Compared with the superficial rectus femoris, peak deoxygenation of the deep rectus femoris (including the superficial intermedius in some) was not significantly different (deoxyhemoglobin and deoxymyoglobin concentration: 81.3 ± 20.8 vs. 78.3 ± 13.6 μM, P > 0.05), but deoxygenation kinetics were significantly slower (mean response time: 37 ± 10 vs. 65 ± 9 s, P ≤ 0.05). These data validate a high-power, time-resolved, near-infrared spectroscopy system with large OS for measuring the deoxygenation of deep tissues and reveal temporal and spatial disparities in muscle deoxygenation responses to exercise. Copyright © 2015 the American Physiological Society.

Image inpainting and super-resolution using non-local recursive deep convolutional network with skip connections

Science.gov (United States)

Liu, Miaofeng

2017-07-01

In recent years, deep convolutional neural networks come into use in image inpainting and super-resolution in many fields. Distinct to most of the former methods requiring to know beforehand the local information for corrupted pixels, we propose a 20-depth fully convolutional network to learn an end-to-end mapping a dataset of damaged/ground truth subimage pairs realizing non-local blind inpainting and super-resolution. As there often exist image with huge corruptions or inpainting on a low-resolution image that the existing approaches unable to perform well, we also share parameters in local area of layers to achieve spatial recursion and enlarge the receptive field. To avoid the difficulty of training this deep neural network, skip-connections between symmetric convolutional layers are designed. Experimental results shows that the proposed method outperforms state-of-the-art methods for diverse corrupting and low-resolution conditions, it works excellently when realizing super-resolution and image inpainting simultaneously

Experimental analysis of two-layered dissimilar metals by roll bonding

Science.gov (United States)

Zhao, Guanghui; Li, Yugui; Li, Juan; Huang, Qingxue; Ma, Lifeng

2018-02-01

Rolling reduction and base layers thickness have important implications for rolling compounding. A two-layered 304 stainless steel/Q345R low alloyed steel was roll bonded. The roll bonding was performed at the three thickness reductions of 25%, 40% and 55% with base layers of various thicknesses (Q345R). The microstructures of the composite were investigated by the ultra-deep microscope (OM) and scanning electron microscope (SEM) and Transmission electron microscope (TEM). Simultaneously, the mechanical properties of the composite were experimentally measured and the tensile fracture surfaces were observed by SEM. The interfaces were successfully bonded without any cracking or voids, which indicated a good fabrication of the 304/Q345R composite. The rolling reduction rate and thinning increase of the substrate contributed to the bonding effects appearance of the roll bonded sheet. The Cr and Ni enriched diffusion layer was formed by the interface elements diffusion. The Cr and Ni diffusion led to the formation of ˜10 μm wide Cr and Ni layers on the carbon steel side.

A 3D Vertically Integrated Deep N-Well CMOS MAPS for the SuperB Layer0

Energy Technology Data Exchange (ETDEWEB)

Traversi, G; Manghisoni, M; Re, V [University of Bergamo, Via Marconi 5, 24044 Dalmine (Italy); Gaioni, L; Ratti, L, E-mail: gianluca.traversi@unibg.it [INFN Pavia, Via Bassi 6, 27100 Pavia (Italy)

2011-01-15

Deep N-Well (DNW) Monolithic Active Pixel Sensors (MAPS) have been developed in the last few years with the aim of building monolithic sensors with similar functionalities as hybrid pixels systems. In these devices the triple well option, available in deep submicron processes, is exploited to implement analog and digital signal processing at the pixel level. Many prototypes have been fabricated in a planar (2D) 130nm CMOS technology. A new kind of DNW-MAPS, namely Apsel5{sub 3}D, which exploits the capabilities of vertical integration (3D) processes, is presented and discussed in this paper. The impact of 3D processes on the design and performance of DNW pixel sensors could be large, with significant advantages in terms of detection efficiency, pixel cell size and immunity to cross-talk, therefore complying with the severe constraints set by future HEP experiments.

On contribution of horizontal and intra-layer convection to the formation of the Baltic Sea cold intermediate layer

Directory of Open Access Journals (Sweden)

I. Chubarenko

2010-02-01

Full Text Available Seasonal cascades down the coastal slopes and intra-layer convection are considered as the two additional mechanisms contributing to the Baltic Sea cold intermediate layer (CIL formation along with conventional seasonal vertical mixing. Field measurements are presented, reporting for the first time the possibility of denser water formation and cascading from the Baltic Sea underwater slopes, which take place under fall and winter cooling conditions and deliver waters into intermediate layer of salinity stratified deep-sea area. The presence in spring within the CIL of water with temperature below that of maximum density (Tmd and that at the local surface in winter time allows tracing its formation: it is argued that the source of the coldest waters of the Baltic CIL is early spring (March–April cascading, arising due to heating of water before reaching the Tmd. Fast increase of the open water heat content during further spring heating indicates that horizontal exchange rather than direct solar heating is responsible for that. When the surface is covered with water, heated above the Tmd, the conditions within the CIL become favorable for intralayer convection due to the presence of waters of Tmd in intermediate layer, which can explain its well-known features – the observed increase of its salinity and deepening with time.

Deep Learning versus Professional Healthcare Equipment: A Fine-Grained Breathing Rate Monitoring Model

Directory of Open Access Journals (Sweden)

Bang Liu

2018-01-01

Full Text Available In mHealth field, accurate breathing rate monitoring technique has benefited a broad array of healthcare-related applications. Many approaches try to use smartphone or wearable device with fine-grained monitoring algorithm to accomplish the task, which can only be done by professional medical equipment before. However, such schemes usually result in bad performance in comparison to professional medical equipment. In this paper, we propose DeepFilter, a deep learning-based fine-grained breathing rate monitoring algorithm that works on smartphone and achieves professional-level accuracy. DeepFilter is a bidirectional recurrent neural network (RNN stacked with convolutional layers and speeded up by batch normalization. Moreover, we collect 16.17 GB breathing sound recording data of 248 hours from 109 and another 10 volunteers to train and test our model, respectively. The results show a reasonably good accuracy of breathing rate monitoring.

Reducing the stair step effect of layer manufactured surfaces by ball burnishing

Science.gov (United States)

Hiegemann, Lars; Agarwal, Chiranshu; Weddeling, Christian; Tekkaya, A. Erman

2016-10-01

The layer technology enables fast and flexible additive manufacturing of forming tools. The disadvantages of this system is the formation of stair steps in the range of tool radii. Within this work a new method to smooth this stair steps by ball burnishing is introduced. This includes studies on the general feasibility of the process and the determination of the influence of the rolling parameters. The investigations are carried out experimentally and numerically. Ultimately, the gained knowledge is applied to finish a deep drawing tool which is manufactured by layer technology.

Characteristics of the Nordic Seas overflows in a set of Norwegian Earth System Model experiments

Science.gov (United States)

Guo, Chuncheng; Ilicak, Mehmet; Bentsen, Mats; Fer, Ilker

2016-08-01

Global ocean models with an isopycnic vertical coordinate are advantageous in representing overflows, as they do not suffer from topography-induced spurious numerical mixing commonly seen in geopotential coordinate models. In this paper, we present a quantitative diagnosis of the Nordic Seas overflows in four configurations of the Norwegian Earth System Model (NorESM) family that features an isopycnic ocean model. For intercomparison, two coupled ocean-sea ice and two fully coupled (atmosphere-land-ocean-sea ice) experiments are considered. Each pair consists of a (non-eddying) 1° and a (eddy-permitting) 1/4° horizontal resolution ocean model. In all experiments, overflow waters remain dense and descend to the deep basins, entraining ambient water en route. Results from the 1/4° pair show similar behavior in the overflows, whereas the 1° pair show distinct differences, including temperature/salinity properties, volume transport (Q), and large scale features such as the strength of the Atlantic Meridional Overturning Circulation (AMOC). The volume transport of the overflows and degree of entrainment are underestimated in the 1° experiments, whereas in the 1/4° experiments, there is a two-fold downstream increase in Q, which matches observations well. In contrast to the 1/4° experiments, the coarse 1° experiments do not capture the inclined isopycnals of the overflows or the western boundary current off the Flemish Cap. In all experiments, the pathway of the Iceland-Scotland Overflow Water is misrepresented: a major fraction of the overflow proceeds southward into the West European Basin, instead of turning westward into the Irminger Sea. This discrepancy is attributed to excessive production of Labrador Sea Water in the model. The mean state and variability of the Nordic Seas overflows have significant consequences on the response of the AMOC, hence their correct representations are of vital importance in global ocean and climate modelling.

Quantum dot laser optimization: selectively doped layers

Science.gov (United States)

Korenev, Vladimir V.; Konoplev, Sergey S.; Savelyev, Artem V.; Shernyakov, Yurii M.; Maximov, Mikhail V.; Zhukov, Alexey E.

2016-08-01

Edge emitting quantum dot (QD) lasers are discussed. It has been recently proposed to use modulation p-doping of the layers that are adjacent to QD layers in order to control QD's charge state. Experimentally it has been proven useful to enhance ground state lasing and suppress the onset of excited state lasing at high injection. These results have been also confirmed with numerical calculations involving solution of drift-diffusion equations. However, deep understanding of physical reasons for such behavior and laser optimization requires analytical approaches to the problem. In this paper, under a set of assumptions we provide an analytical model that explains major effects of selective p-doping. Capture rates of elections and holes can be calculated by solving Poisson equations for electrons and holes around the charged QD layer. The charge itself is ruled by capture rates and selective doping concentration. We analyzed this self-consistent set of equations and showed that it can be used to optimize QD laser performance and to explain underlying physics.

Quantum dot laser optimization: selectively doped layers

International Nuclear Information System (INIS)

Korenev, Vladimir V; Konoplev, Sergey S; Savelyev, Artem V; Shernyakov, Yurii M; Maximov, Mikhail V; Zhukov, Alexey E

2016-01-01

Edge emitting quantum dot (QD) lasers are discussed. It has been recently proposed to use modulation p-doping of the layers that are adjacent to QD layers in order to control QD's charge state. Experimentally it has been proven useful to enhance ground state lasing and suppress the onset of excited state lasing at high injection. These results have been also confirmed with numerical calculations involving solution of drift-diffusion equations. However, deep understanding of physical reasons for such behavior and laser optimization requires analytical approaches to the problem. In this paper, under a set of assumptions we provide an analytical model that explains major effects of selective p-doping. Capture rates of elections and holes can be calculated by solving Poisson equations for electrons and holes around the charged QD layer. The charge itself is ruled by capture rates and selective doping concentration. We analyzed this self-consistent set of equations and showed that it can be used to optimize QD laser performance and to explain underlying physics. (paper)

Deep Subseafloor Fungi as an Untapped Reservoir of Amphipathic Antimicrobial Compounds.

Science.gov (United States)

Navarri, Marion; Jégou, Camille; Meslet-Cladière, Laurence; Brillet, Benjamin; Barbier, Georges; Burgaud, Gaëtan; Fleury, Yannick

2016-03-10

The evolving global threat of antimicrobial resistance requires a deep renewal of the antibiotic arsenal including the isolation and characterization of new drugs. Underexplored marine ecosystems may represent an untapped reservoir of novel bioactive molecules. Deep-sea fungi isolated from a record-depth sediment core of almost 2000 m below the seafloor were investigated for antimicrobial activities. This antimicrobial screening, using 16 microbial targets, revealed 33% of filamentous fungi synthesizing bioactive compounds with activities against pathogenic bacteria and fungi. Interestingly, occurrence of antimicrobial producing isolates was well correlated with the complexity of the habitat (in term of microbial richness), as higher antimicrobial activities were obtained at specific layers of the sediment core. It clearly highlights complex deep-sea habitats as chemical battlefields where synthesis of numerous bioactive compounds appears critical for microbial competition. The six most promising deep subseafloor fungal isolates were selected for the production and extraction of bioactive compounds. Depending on the fungal isolates, antimicrobial compounds were only biosynthesized in semi-liquid or solid-state conditions as no antimicrobial activities were ever detected using liquid fermentation. An exception was made for one fungal isolate, and the extraction procedure designed to extract amphipathic compounds was successful and highlighted the amphiphilic profile of the bioactive metabolites.

On the Reduction of Computational Complexity of Deep Convolutional Neural Networks

Directory of Open Access Journals (Sweden)

Partha Maji

2018-04-01

Full Text Available Deep convolutional neural networks (ConvNets, which are at the heart of many new emerging applications, achieve remarkable performance in audio and visual recognition tasks. Unfortunately, achieving accuracy often implies significant computational costs, limiting deployability. In modern ConvNets it is typical for the convolution layers to consume the vast majority of computational resources during inference. This has made the acceleration of these layers an important research area in academia and industry. In this paper, we examine the effects of co-optimizing the internal structures of the convolutional layers and underlying implementation of fundamental convolution operation. We demonstrate that a combination of these methods can have a big impact on the overall speedup of a ConvNet, achieving a ten-fold increase over baseline. We also introduce a new class of fast one-dimensional (1D convolutions for ConvNets using the Toom–Cook algorithm. We show that our proposed scheme is mathematically well-grounded, robust, and does not require any time-consuming retraining, while still achieving speedups solely from convolutional layers with no loss in baseline accuracy.

Omni-directional reflectors for deep blue LED using symmetric autocloning method

Science.gov (United States)

Chen, Sheng-Hui; Chen, Chun-Ko; Huang, Yu-Chia; Lee, Cheng-Chung

2013-03-01

Omni-directional reflectors (ODRs) for deep blue LED were designed and fabricated using symmetric autocloning method. The symmetric stack multi-layers for the reflectors were designed by finite-difference time-domain simulation. The fabricating process of ODR is combined with the techniques of anodic aluminum oxide (AAO) process and autocloning method. The two-dimensional structure template of nano-channel array was grown using AAO with the period of 150 nm. Then the shaping layer was deposited on the AAO template by evaporation deposition. Besides, the ion etching was applied to modify the apex angle to the triangle shape at 100°. Finally, the sub/(0.5TiO2 SiO2 0.5TiO2)8 multi-layer stack was deposited on the shaping layer using autocloning method to achieve the ODR. The results show the reflective spectra of ODR at the incident angles of 0, 30, 45, and 60° had high values within the range 400-450 nm. Besides, the central wavelength shifting is not obvious which is very good for keeping the color of LED stable.

Spiraling pathways of global deep waters to the surface of the Southern Ocean.

Science.gov (United States)

Tamsitt, Veronica; Drake, Henri F; Morrison, Adele K; Talley, Lynne D; Dufour, Carolina O; Gray, Alison R; Griffies, Stephen M; Mazloff, Matthew R; Sarmiento, Jorge L; Wang, Jinbo; Weijer, Wilbert

2017-08-02

Upwelling of global deep waters to the sea surface in the Southern Ocean closes the global overturning circulation and is fundamentally important for oceanic uptake of carbon and heat, nutrient resupply for sustaining oceanic biological production, and the melt rate of ice shelves. However, the exact pathways and role of topography in Southern Ocean upwelling remain largely unknown. Here we show detailed upwelling pathways in three dimensions, using hydrographic observations and particle tracking in high-resolution models. The analysis reveals that the northern-sourced deep waters enter the Antarctic Circumpolar Current via southward flow along the boundaries of the three ocean basins, before spiraling southeastward and upward through the Antarctic Circumpolar Current. Upwelling is greatly enhanced at five major topographic features, associated with vigorous mesoscale eddy activity. Deep water reaches the upper ocean predominantly south of the Antarctic Circumpolar Current, with a spatially nonuniform distribution. The timescale for half of the deep water to upwell from 30° S to the mixed layer is ~60-90 years.Deep waters of the Atlantic, Pacific and Indian Oceans upwell in the Southern Oceanbut the exact pathways are not fully characterized. Here the authors present a three dimensional view showing a spiralling southward path, with enhanced upwelling by eddy-transport at topographic hotspots.

Microbially induced corrosion of carbon steel in deep groundwater environment

Directory of Open Access Journals (Sweden)

Pauliina eRajala

2015-07-01

Full Text Available The metallic low and intermediate level radioactive waste generally consists of carbon steel and stainless steels. The corrosion rate of carbon steel in deep groundwater is typically low, unless the water is very acidic or microbial activity in the environment is high. Therefore, the assessment of microbially induced corrosion of carbon steel in deep bedrock environment has become important for evaluating the safety of disposal of radioactive waste. Here we studied the corrosion inducing ability of indigenous microbial community from a deep bedrock aquifer. Carbon steel coupons were exposed to anoxic groundwater from repository site 100 m depth (Olkiluoto, Finland for periods of three and eight months. The experiments were conducted at both in situ temperature and room temperature to investigate the response of microbial population to elevated temperature. Our results demonstrate that microorganisms from the deep bedrock aquifer benefit from carbon steel introduced to the nutrient poor anoxic deep groundwater environment. In the groundwater incubated with carbon steel the planktonic microbial community was more diverse and 100-fold more abundant compared to the environment without carbon steel. The betaproteobacteria were the most dominant bacterial class in all samples where carbon steel was present, whereas in groundwater incubated without carbon steel the microbial community had clearly less diversity. Microorganisms induced pitting corrosion and were found to cluster inside the corrosion pits. Temperature had an effect on the species composition of microbial community and also affected the corrosion deposits layer formed on the surface of carbon steel.

Deep and intermediate mediterranean water in the western Alboran Sea

Science.gov (United States)

Parrilla, Gregorio; Kinder, Thomas H.; Preller, Ruth H.

1986-01-01

Hydrographic and current meter data, obtained during June to October 1982, and numerical model experiments are used to study the distribution and flow of Mediterranean waters in the western Alboran Sea. The Intermediate Water is more pronounced in the northern three-fourths of the sea, but its distribution is patchy as manifested by variability of the temperature and salinity maxima at scales ≤10 km. Current meters in the lower Intermediate Water showed mean flow toward the Strait at 2 cm s -1. A reversal of this flow lasted about 2 weeks. A rough estimate of the mean westward Intermediate Water transport was 0.4 × 10 6 m 3 s -1, about one-third of the total outflow, so that the best estimates of the contributions of traditionally defined Intermediate Water and Deep Water account for only about one-half of the total outflow. The Deep Water was uplifted against the southern continental slope from Alboran Island (3°W) to the Strait. There was also a similar but much weaker banking against the Spanish slope, but a deep current record showed that the eastward recirculation implied by this banking is probably intermittent. Two-layer numerical model experiments simulated the Intermediate Water flow with a flat bottom and the Deep Water with realistic bottom topography. Both experiments replicated the major circulation features, and the Intermediate Water flow was concentrated in the north because of rotation and the Deep Water flow in the south because of topographic control.

Deep transfer learning for automatic target classification: MWIR to LWIR

Science.gov (United States)

Ding, Zhengming; Nasrabadi, Nasser; Fu, Yun

2016-05-01

Publisher's Note: This paper, originally published on 5/12/2016, was replaced with a corrected/revised version on 5/18/2016. If you downloaded the original PDF but are unable to access the revision, please contact SPIE Digital Library Customer Service for assistance. When dealing with sparse or no labeled data in the target domain, transfer learning shows its appealing performance by borrowing the supervised knowledge from external domains. Recently deep structure learning has been exploited in transfer learning due to its attractive power in extracting effective knowledge through multi-layer strategy, so that deep transfer learning is promising to address the cross-domain mismatch. In general, cross-domain disparity can be resulted from the difference between source and target distributions or different modalities, e.g., Midwave IR (MWIR) and Longwave IR (LWIR). In this paper, we propose a Weighted Deep Transfer Learning framework for automatic target classification through a task-driven fashion. Specifically, deep features and classifier parameters are obtained simultaneously for optimal classification performance. In this way, the proposed deep structures can extract more effective features with the guidance of the classifier performance; on the other hand, the classifier performance is further improved since it is optimized on more discriminative features. Furthermore, we build a weighted scheme to couple source and target output by assigning pseudo labels to target data, therefore we can transfer knowledge from source (i.e., MWIR) to target (i.e., LWIR). Experimental results on real databases demonstrate the superiority of the proposed algorithm by comparing with others.

Physical properties of a new Deep Eutectic Solvent based on lithium bis[(trifluoromethyl)sulfonyl]imide and N-methylacetamide as superionic suitable electrolyte for lithium ion batteries and electric double layer capacitors

International Nuclear Information System (INIS)

Boisset, Aurélien; Jacquemin, Johan; Anouti, Mérièm

2013-01-01

Highlights: • Preparation of new Deep Eutectic Solvent (DES) based on N-methylacetamide and TFSI. • Characterization of conductivity, viscosity and thermal properties of DES. • DES presents a superionic character in Walden classification. • DES is suitable electrolyte for lithium ion batteries and electric double layer capacitors. -- Abstract: Herein we present a study on the physical/chemical properties of a new Deep Eutectic Solvent (DES) based on N-methylacetamide (MAc) and lithium bis[(trifluoromethyl)sulfonyl]imide (LiTFSI). Due to its interesting properties, such as wide liquid-phase range from −60 °C to 280 °C, low vapor pressure, and high ionic conductivity up to 28.4 mS cm −1 at 150 °C and at x LiTFSI = 1/4, this solution can be practically used as electrolyte for electrochemical storage systems such as electric double-layer capacitors (EDLCs) and/or lithium ion batteries (LiBs). Firstly, relationships between its transport properties (conductivity and viscosity) as a function of composition and temperature were discussed through Arrhenius’ Law and Vogel–Tamman–Fulcher (VTF) equations, as well as by using the Walden classification. From this investigation, it appears that this complex electrolyte possesses a number of excellent transport properties, like a superionic character for example. Based on which, we then evaluated its electrochemical performances as electrolyte for EDLCs and LiBs applications by using activated carbon (AC) and lithium iron phosphate (LiFePO 4 ) electrodes, respectively. These results demonstrate that this electrolyte has a good compatibility with both electrodes (AC and LiFePO 4 ) in each testing cell driven also by excellent electrochemical properties in specific capacitance, rate and cycling performances, indicating that the LiTFSI/MAc DES can be a promising electrolyte for EDLCs and LiBs applications especially for those requiring high safety and stability

Deep neural network convolution (NNC) for three-class classification of diffuse lung disease opacities in high-resolution CT (HRCT): consolidation, ground-glass opacity (GGO), and normal opacity

Science.gov (United States)

Hashimoto, Noriaki; Suzuki, Kenji; Liu, Junchi; Hirano, Yasushi; MacMahon, Heber; Kido, Shoji

2018-02-01

Consolidation and ground-glass opacity (GGO) are two major types of opacities associated with diffuse lung diseases. Accurate detection and classification of such opacities are crucially important in the diagnosis of lung diseases, but the process is subjective, and suffers from interobserver variability. Our study purpose was to develop a deep neural network convolution (NNC) system for distinguishing among consolidation, GGO, and normal lung tissue in high-resolution CT (HRCT). We developed ensemble of two deep NNC models, each of which was composed of neural network regression (NNR) with an input layer, a convolution layer, a fully-connected hidden layer, and a fully-connected output layer followed by a thresholding layer. The output layer of each NNC provided a map for the likelihood of being each corresponding lung opacity of interest. The two NNC models in the ensemble were connected in a class-selection layer. We trained our NNC ensemble with pairs of input 2D axial slices and "teaching" probability maps for the corresponding lung opacity, which were obtained by combining three radiologists' annotations. We randomly selected 10 and 40 slices from HRCT scans of 172 patients for each class as a training and test set, respectively. Our NNC ensemble achieved an area under the receiver-operating-characteristic (ROC) curve (AUC) of 0.981 and 0.958 in distinction of consolidation and GGO, respectively, from normal opacity, yielding a classification accuracy of 93.3% among 3 classes. Thus, our deep-NNC-based system for classifying diffuse lung diseases achieved high accuracies for classification of consolidation, GGO, and normal opacity.

LookSeq: A browser-based viewer for deep sequencing data

OpenAIRE

Manske, Heinrich Magnus; Kwiatkowski, Dominic P.

2009-01-01

Sequencing a genome to great depth can be highly informative about heterogeneity within an individual or a population. Here we address the problem of how to visualize the multiple layers of information contained in deep sequencing data. We propose an interactive AJAX-based web viewer for browsing large data sets of aligned sequence reads. By enabling seamless browsing and fast zooming, the LookSeq program assists the user to assimilate information at different levels of resolution, from an ov...

A Multiobjective Sparse Feature Learning Model for Deep Neural Networks.

Science.gov (United States)

Gong, Maoguo; Liu, Jia; Li, Hao; Cai, Qing; Su, Linzhi

2015-12-01

Hierarchical deep neural networks are currently popular learning models for imitating the hierarchical architecture of human brain. Single-layer feature extractors are the bricks to build deep networks. Sparse feature learning models are popular models that can learn useful representations. But most of those models need a user-defined constant to control the sparsity of representations. In this paper, we propose a multiobjective sparse feature learning model based on the autoencoder. The parameters of the model are learnt by optimizing two objectives, reconstruction error and the sparsity of hidden units simultaneously to find a reasonable compromise between them automatically. We design a multiobjective induced learning procedure for this model based on a multiobjective evolutionary algorithm. In the experiments, we demonstrate that the learning procedure is effective, and the proposed multiobjective model can learn useful sparse features.

Pathways of upwelling deep waters to the surface of the Southern Ocean

Science.gov (United States)

Tamsitt, Veronica; Drake, Henri; Morrison, Adele; Talley, Lynne; Dufour, Carolina; Gray, Alison; Griffies, Stephen; Mazloff, Matthew; Sarmiento, Jorge; Wang, Jinbo; Weijer, Wilbert

2017-04-01

Upwelling of Atlantic, Indian and Pacific deep waters to the sea surface in the Southern Ocean closes the global overturning circulation and is fundamentally important for oceanic uptake of anthropogenic carbon and heat, nutrient resupply for sustaining oceanic biological production, and the melt rate of ice shelves. Here we go beyond the two-dimensional view of Southern Ocean upwelling, to show detailed Southern Ocean upwelling pathways in three dimensions, using hydrographic observations and particle tracking in high-resolution ocean and climate models. The northern deep waters enter the Antarctic Circumpolar Current (ACC) via narrow southward currents along the boundaries of the three ocean basins, before spiraling southeastward and upward through the ACC. Upwelling is greatly enhanced at five major topographic features, associated with vigorous mesoscale eddy activity. Deep water reaches the upper ocean predominantly south of the southern ACC boundary, with a spatially nonuniform distribution, regionalizing warm water supply to Antarctic ice shelves and the delivery of nutrient and carbon-rich water to the sea surface. The timescale for half of the deep water to upwell from 30°S to the mixed layer is on the order of 60-90 years, which has important implications for the timescale for signals to propagate through the deep ocean. In addition, we quantify the diabatic transformation along particle trajectories, to identify where diabatic processes are important along the upwelling pathways.

A novel deep learning approach for classification of EEG motor imagery signals.

Science.gov (United States)

Tabar, Yousef Rezaei; Halici, Ugur

2017-02-01

Signal classification is an important issue in brain computer interface (BCI) systems. Deep learning approaches have been used successfully in many recent studies to learn features and classify different types of data. However, the number of studies that employ these approaches on BCI applications is very limited. In this study we aim to use deep learning methods to improve classification performance of EEG motor imagery signals. In this study we investigate convolutional neural networks (CNN) and stacked autoencoders (SAE) to classify EEG Motor Imagery signals. A new form of input is introduced to combine time, frequency and location information extracted from EEG signal and it is used in CNN having one 1D convolutional and one max-pooling layers. We also proposed a new deep network by combining CNN and SAE. In this network, the features that are extracted in CNN are classified through the deep network SAE. The classification performance obtained by the proposed method on BCI competition IV dataset 2b in terms of kappa value is 0.547. Our approach yields 9% improvement over the winner algorithm of the competition. Our results show that deep learning methods provide better classification performance compared to other state of art approaches. These methods can be applied successfully to BCI systems where the amount of data is large due to daily recording.

The hydrography of the Mozambique Channel from six years of continuous temperature, salinity, and velocity observations

Science.gov (United States)

Ullgren, J. E.; van Aken, H. M.; Ridderinkhof, H.; de Ruijter, W. P. M.

2012-11-01

Temperature, salinity and velocity data are presented, along with the estimated volume transport, from seven full-length deep sea moorings placed across the narrowest part of the Mozambique Channel, southwest Indian Ocean, during the period November 2003 to December 2009. The dominant water mass in the upper layer is Sub-Tropical Surface Water (STSW) which overlies South Indian Central Water (SICW), and is normally capped by fresher Tropical Surface Water (TSW). Upper ocean salinity increased through 2005 as a result of saline STSW taking up a relatively larger part of the upper layer, at the expense of TSW. Upper waters are on average warmer and lighter in the central Channel than on the sides. Throughout the upper 1.5 km of the water column there is large hydrographic variability, short-term as well as interannual, and in particular at frequencies (four to seven cycles per year) associated with the southward passage of anticyclonic Mozambique Channel eddies. The eddies have a strong T-S signal, in the upper and central waters as well as on the intermediate level, as the eddies usually carry saline Red Sea Water (RSW) in their core. While the interannual frequency band displays an east-west gradient with higher temperature variance on the western side, the eddy frequency band shows highest variance in the centre of the Channel, where the eddy band contains about 40% of the total isopycnal hydrographic variability. Throughout the >6 years of measurements, the frequency and characteristics of eddies vary between periods, both in terms of strength and vertical structure of eddy T-S signals. These changes contribute to the interannual variability of water mass properties: an increase in central water salinity to a maximum in late 2007 coincided with a period of unusually frequent eddies with strong salinity signals. The warmest and most saline deep water is found within the northward flowing Mozambique Undercurrent, on the western side of the Channel. The Undercurrent

A novel biomedical image indexing and retrieval system via deep preference learning.

Science.gov (United States)

Pang, Shuchao; Orgun, Mehmet A; Yu, Zhezhou

2018-05-01

The traditional biomedical image retrieval methods as well as content-based image retrieval (CBIR) methods originally designed for non-biomedical images either only consider using pixel and low-level features to describe an image or use deep features to describe images but still leave a lot of room for improving both accuracy and efficiency. In this work, we propose a new approach, which exploits deep learning technology to extract the high-level and compact features from biomedical images. The deep feature extraction process leverages multiple hidden layers to capture substantial feature structures of high-resolution images and represent them at different levels of abstraction, leading to an improved performance for indexing and retrieval of biomedical images. We exploit the current popular and multi-layered deep neural networks, namely, stacked denoising autoencoders (SDAE) and convolutional neural networks (CNN) to represent the discriminative features of biomedical images by transferring the feature representations and parameters of pre-trained deep neural networks from another domain. Moreover, in order to index all the images for finding the similarly referenced images, we also introduce preference learning technology to train and learn a kind of a preference model for the query image, which can output the similarity ranking list of images from a biomedical image database. To the best of our knowledge, this paper introduces preference learning technology for the first time into biomedical image retrieval. We evaluate the performance of two powerful algorithms based on our proposed system and compare them with those of popular biomedical image indexing approaches and existing regular image retrieval methods with detailed experiments over several well-known public biomedical image databases. Based on different criteria for the evaluation of retrieval performance, experimental results demonstrate that our proposed algorithms outperform the state

On modelling of lateral buckling failure in flexible pipe tensile armour layers

DEFF Research Database (Denmark)

Østergaard, Niels Højen; Lyckegaard, Anders; Andreasen, Jens H.

2012-01-01

In the present paper, a mathematical model which is capable of representing the physics of lateral buckling failure in the tensile armour layers of flexible pipes is introduced. Flexible pipes are unbounded composite steel–polymer structures, which are known to be prone to lateral wire buckling...... when exposed to repeated bending cycles and longitudinal compression, which mainly occurs during pipe laying in ultra-deep waters. On the basis of multiple single wire analyses, the mechanical behaviour of both layers of tensile armour wires can be determined. Since failure in one layer destabilises...... the torsional equilibrium which is usually maintained between the layers, lateral wire buckling is often associated with a severe pipe twist. This behaviour is discussed and modelled. Results are compared to a pipe model, in which failure is assumed not to cause twist. The buckling modes of the tensile armour...

Engineering Seismic Base Layer for Defining Design Earthquake Motion

International Nuclear Information System (INIS)

Yoshida, Nozomu

2008-01-01

Engineer's common sense that incident wave is common in a widespread area at the engineering seismic base layer is shown not to be correct. An exhibiting example is first shown, which indicates that earthquake motion at the ground surface evaluated by the analysis considering the ground from a seismic bedrock to a ground surface simultaneously (continuous analysis) is different from the one by the analysis in which the ground is separated at the engineering seismic base layer and analyzed separately (separate analysis). The reason is investigated by several approaches. Investigation based on eigen value problem indicates that the first predominant period in the continuous analysis cannot be found in the separate analysis, and predominant period at higher order does not match in the upper and lower ground in the separate analysis. The earthquake response analysis indicates that reflected wave at the engineering seismic base layer is not zero, which indicates that conventional engineering seismic base layer does not work as expected by the term ''base''. All these results indicate that wave that goes down to the deep depths after reflecting in the surface layer and again reflects at the seismic bedrock cannot be neglected in evaluating the response at the ground surface. In other words, interaction between the surface layer and/or layers between seismic bedrock and engineering seismic base layer cannot be neglected in evaluating the earthquake motion at the ground surface

Characterization of deep level defects and thermally stimulated depolarization phenomena in La-doped TlInS2 layered semiconductor

International Nuclear Information System (INIS)

Seyidov, MirHasan Yu.; Suleymanov, Rauf A.; Mikailzade, Faik A.; Kargın, Elif Orhan; Odrinsky, Andrei P.

2015-01-01

Lanthanum-doped high quality TlInS 2 (TlInS 2 :La) ferroelectric-semiconductor was characterized by photo-induced current transient spectroscopy (PICTS). Different impurity centers are resolved and identified. Analyses of the experimental data were performed in order to determine the characteristic parameters of the extrinsic and intrinsic defects. The energies and capturing cross section of deep traps were obtained by using the heating rate method. The observed changes in the Thermally Stimulated Depolarization Currents (TSDC) near the phase transition points in TlInS 2 :La ferroelectric-semiconductor are interpreted as a result of self-polarization of the crystal due to the internal electric field caused by charged defects. The TSDC spectra show the depolarization peaks, which are attributed to defects of dipolar origin. These peaks provide important information on the defect structure and localized energy states in TlInS 2 :La. Thermal treatments of TlInS 2 :La under an external electric field, which was applied at different temperatures, allowed us to identify a peak in TSDC which was originated from La-dopant. It was established that deep energy level trap BTE43, which are active at low temperature (T ≤ 156 K) and have activation energy 0.29 eV and the capture cross section 2.2 × 10 −14 cm 2 , corresponds to the La dopant. According to the PICTS results, the deep level trap center B5 is activated in the temperature region of incommensurate (IC) phases of TlInS 2 :La, having the giant static dielectric constant due to the structural disorders. From the PICTS simulation results for B5, native deep level trap having an activation energy of 0.3 eV and the capture cross section of 1.8 × 10 −16 cm 2 were established. A substantial amount of residual space charges is trapped by the deep level localized energy states of B5 in IC-phase. While the external electric field is applied, permanent dipoles, which are originated from the charged B5

Biosphere modelling for a deep radioactive waste repository: treatment of the groundwater-soil pathway

International Nuclear Information System (INIS)

Baeyens, B.; Grogan, H.A.; Dorp, F. van

1991-07-01

The effect of radionuclide transfer from near-surface groundwater to the rooting zone soil, via a deep soil layer, is modelled in this report. The possible extent of upward solute movement is evaluated for a region in northern Switzerland. The concentration of 237 Np and 129 I in the deep and top soil, and hence growing crops, are evaluated assuming a constant unit activity concentration in the groundwater. A number of parameter variations are considered, namely variable soil sorption coefficients, reduced infiltration of rain water and decreased groundwater flow. A release to an alternative smaller recipient region in northern Switzerland is also evaluated. For the parameter ranges considered uncertainty in the solid-liquid distribution coefficient has the largest effect on overall uncertainty. These calculations have been presented within the international Biosphere Model Validation Study (BIOMOVS). A description of the test scenario, and the model calculations submitted, have been included in this report for completeness. To place the groundwater-soil-crop-man pathway in context, its contribution to the total dose to man is evaluated for the 237 Np- 233 U- 229 Th decay chain. The results obtained using the two-layer soil model, described in this report, are compared with the single-layer soil model used during Project Gewaehr 1985. The more realistic two-layer soil model indicated an increase in importance of the drinking water pathway. It should be noted, however, that not all the critical pathways have been treated in this study with the same degree of conservatism. (author) 16 figs., 15 tabs., 31 refs

Training Deep Convolutional Neural Networks with Resistive Cross-Point Devices

Directory of Open Access Journals (Sweden)

Tayfun Gokmen

2017-10-01

Full Text Available In a previous work we have detailed the requirements for obtaining maximal deep learning performance benefit by implementing fully connected deep neural networks (DNN in the form of arrays of resistive devices. Here we extend the concept of Resistive Processing Unit (RPU devices to convolutional neural networks (CNNs. We show how to map the convolutional layers to fully connected RPU arrays such that the parallelism of the hardware can be fully utilized in all three cycles of the backpropagation algorithm. We find that the noise and bound limitations imposed by the analog nature of the computations performed on the arrays significantly affect the training accuracy of the CNNs. Noise and bound management techniques are presented that mitigate these problems without introducing any additional complexity in the analog circuits and that can be addressed by the digital circuits. In addition, we discuss digitally programmable update management and device variability reduction techniques that can be used selectively for some of the layers in a CNN. We show that a combination of all those techniques enables a successful application of the RPU concept for training CNNs. The techniques discussed here are more general and can be applied beyond CNN architectures and therefore enables applicability of the RPU approach to a large class of neural network architectures.

Ploughing in simulated radioactive layer on farming solid

International Nuclear Information System (INIS)

Nilsson, J.

1983-09-01

In case of a nuclear disaster, farmland would be contaminated by nuclear fallout. The possibilities of restoring such land by different means is under investigation at the Department of Radioecology. One of these means is mouldboard ploughing for placing a contaminated surface layer deep in comparison to the subsequent soil managment operations. The placement efficiencies obtained under varying experimental conditions are given in this report. Different widths of cut, working depths and driving speeds have been studied as well as different jointers, a foreplough and a trash board. A double-depth plough, a -parallellplough- and a plough with a deep-digging body were included in the test. Large cutting width and great working depth proved benficial, while neither use jointers, foreplough, trash board nor changes in driving speed had any apparent effect on the result of the ploughing. The double-depth plough worked as well as a conventional plough with the same width of cut, while the -parallellplough- did not perform well. The best results were experienced with the deep-digging body (cutting width 600 mm, 24) when used at great working depth (500 mm, 20). This type of plough is, however, rare in Sweden. (author)

Automatic Classification of volcano-seismic events based on Deep Neural Networks.

Science.gov (United States)

Titos Luzón, M.; Bueno Rodriguez, A.; Garcia Martinez, L.; Benitez, C.; Ibáñez, J. M.

2017-12-01

Seismic monitoring of active volcanoes is a popular remote sensing technique to detect seismic activity, often associated to energy exchanges between the volcano and the environment. As a result, seismographs register a wide range of volcano-seismic signals that reflect the nature and underlying physics of volcanic processes. Machine learning and signal processing techniques provide an appropriate framework to analyze such data. In this research, we propose a new classification framework for seismic events based on deep neural networks. Deep neural networks are composed by multiple processing layers, and can discover intrinsic patterns from the data itself. Internal parameters can be initialized using a greedy unsupervised pre-training stage, leading to an efficient training of fully connected architectures. We aim to determine the robustness of these architectures as classifiers of seven different types of seismic events recorded at "Volcán de Fuego" (Colima, Mexico). Two deep neural networks with different pre-training strategies are studied: stacked denoising autoencoder and deep belief networks. Results are compared to existing machine learning algorithms (SVM, Random Forest, Multilayer Perceptron). We used 5 LPC coefficients over three non-overlapping segments as training features in order to characterize temporal evolution, avoid redundancy and encode the signal, regardless of its duration. Experimental results show that deep architectures can classify seismic events with higher accuracy than classical algorithms, attaining up to 92% recognition accuracy. Pre-training initialization helps these models to detect events that occur simultaneously in time (such explosions and rockfalls), increase robustness against noisy inputs, and provide better generalization. These results demonstrate deep neural networks are robust classifiers, and can be deployed in real-environments to monitor the seismicity of restless volcanoes.

Effect of deep structure and surface layer (a) simulation study on the heavy damage belt of the 1995 Hyogoken Nanbu earthquake; Hado simulation wo mochiita `shinsai no obi` ni tsuite no kento

Energy Technology Data Exchange (ETDEWEB)

Feng, S; Ishikawa, K; Kaji, Y [Chuokaihatsu Corp., Tokyo (Japan)

1996-05-01

A simulation study was done to identify the causes for the so-called heavy damage belt observed as a result of the Hyogoken Nanbu Earthquake. This study determines distributions of the maximum amplitudes on the ground surfaces, and discusses the effects of deep structures and low-velocity surface layers, based on the simulation by the wave equation with the underground model of Higashinada-ku and its vicinity in the north-south direction, observed seismic records and artificial waves. The two-dimensional scalar wave equation is used for the analysis. The velocity structure model used for the simulation is established, based on the elastic wave seismic survey results. The focus function is drawn by expanding or contracting the time scale, using the seismic records at Kobe Port Island and artificial waves. The analysis results show that the damage belt coincides with the areas at which the focusing zone of the deep structure overlap the amplification zone in the low-velocity ground surfaces, where relative density is amplified 1.5 to 2 times. It is also observed that large peaks repeat 2 to 3 times on the time scale. 5 refs., 8 figs., 1 tab.

SYNAPTIC DEPRESSION IN DEEP NEURAL NETWORKS FOR SPEECH PROCESSING.

Science.gov (United States)

Zhang, Wenhao; Li, Hanyu; Yang, Minda; Mesgarani, Nima

2016-03-01

A characteristic property of biological neurons is their ability to dynamically change the synaptic efficacy in response to variable input conditions. This mechanism, known as synaptic depression, significantly contributes to the formation of normalized representation of speech features. Synaptic depression also contributes to the robust performance of biological systems. In this paper, we describe how synaptic depression can be modeled and incorporated into deep neural network architectures to improve their generalization ability. We observed that when synaptic depression is added to the hidden layers of a neural network, it reduces the effect of changing background activity in the node activations. In addition, we show that when synaptic depression is included in a deep neural network trained for phoneme classification, the performance of the network improves under noisy conditions not included in the training phase. Our results suggest that more complete neuron models may further reduce the gap between the biological performance and artificial computing, resulting in networks that better generalize to novel signal conditions.

Astronomical calibration of the first Toba super-eruption from deep-sea sediments

Science.gov (United States)

Lee, M.; Chen, C.; Wei, K.; Iizuka, Y.

2003-04-01

Correlations between tephra layers interbedded within deep-sea cores and radiometrically dated volcanic eruptions provide an independent means of verifying dating techniques developed for sediment cores. Alternatively, the chronostratigraphic framework developed from marine sediments can be used to calibrate ages of land-base eruptions, if geochemical correlations can be established. In this study, we examined three deep-sea cores along an east-west transection across the South China Sea, with a distance of ~1800 to 2500 km away from the Toba caldera. The occurrence of the Oldest Toba Tuff was recognized on the basis of its geochemical characteristics, such as a high-silicate, high-potassium content and a distinct strontium isotope composition. The correlative tephra layer occurs slightly above the Australasian microtektite layer and below the Brunhes/Matuyama boundary, which in constitute three time-parallel markers for correlation and dating of Quaternary stratigraphic records. Against the astronomically tuned oxygen isotope chronostratigraphy, the rhyolitic ignimbrite erupted during the transition from marine isotope stage 20 (glacial) to stage 19 (interglacial) with an estimated age of 788 ka. The refined age is in good agreement with the radiometric age of 800+20 ka for Layer D of ODP Site 758 (Hall and Farrell, 1995), but significantly younger than the commonly referred age of 840+30 ka (Diehl et al., 1987). The mid-Pleistocene eruption expelled at least 800-1000 km3 dense-rock-equivalent of rhyolitic magma taking into account the widespread ashfall deposits in the Indian Ocean and the South China Sea basins. In spite of its exceptional magnitude, the timing of the first Toba super-eruption disputes a possible causal linkage between a major volcanic eruption and a long-term global climatic deterioration.

Prediction of Bispectral Index during Target-controlled Infusion of Propofol and Remifentanil: A Deep Learning Approach.

Science.gov (United States)

Lee, Hyung-Chul; Ryu, Ho-Geol; Chung, Eun-Jin; Jung, Chul-Woo

2018-03-01

The discrepancy between predicted effect-site concentration and measured bispectral index is problematic during intravenous anesthesia with target-controlled infusion of propofol and remifentanil. We hypothesized that bispectral index during total intravenous anesthesia would be more accurately predicted by a deep learning approach. Long short-term memory and the feed-forward neural network were sequenced to simulate the pharmacokinetic and pharmacodynamic parts of an empirical model, respectively, to predict intraoperative bispectral index during combined use of propofol and remifentanil. Inputs of long short-term memory were infusion histories of propofol and remifentanil, which were retrieved from target-controlled infusion pumps for 1,800 s at 10-s intervals. Inputs of the feed-forward network were the outputs of long short-term memory and demographic data such as age, sex, weight, and height. The final output of the feed-forward network was the bispectral index. The performance of bispectral index prediction was compared between the deep learning model and previously reported response surface model. The model hyperparameters comprised 8 memory cells in the long short-term memory layer and 16 nodes in the hidden layer of the feed-forward network. The model training and testing were performed with separate data sets of 131 and 100 cases. The concordance correlation coefficient (95% CI) were 0.561 (0.560 to 0.562) in the deep learning model, which was significantly larger than that in the response surface model (0.265 [0.263 to 0.266], P deep learning model-predicted bispectral index during target-controlled infusion of propofol and remifentanil more accurately compared to the traditional model. The deep learning approach in anesthetic pharmacology seems promising because of its excellent performance and extensibility.

Theoretical simulation of the dual-heat-flux method in deep body temperature measurements.

Science.gov (United States)

Huang, Ming; Chen, Wenxi

2010-01-01

Deep body temperature reveals individual physiological states, and is important in patient monitoring and chronobiological studies. An innovative dual-heat-flux method has been shown experimentally to be competitive with the conventional zero-heat-flow method in its performance, in terms of measurement accuracy and step response to changes in the deep temperature. We have utilized a finite element method to model and simulate the dynamic process of a dual-heat-flux probe in deep body temperature measurements to validate the fundamental principles of the dual-heat-flux method theoretically, and to acquire a detailed quantitative description of the thermal profile of the dual-heat-flux probe. The simulation results show that the estimated deep body temperature is influenced by the ambient temperature (linearly, at a maximum rate of 0.03 °C/°C) and the blood perfusion rate. The corresponding depth of the estimated temperature in the skin and subcutaneous tissue layer is consistent when using the dual-heat-flux probe. Insights in improving the performance of the dual-heat-flux method were discussed for further studies of dual-heat-flux probes, taking into account structural and geometric considerations.

Performance comparison of CO2 and diode lasers for deep-section concrete cutting

International Nuclear Information System (INIS)

Crouse, Philip L.; Li, Lin; Spencer, Julian T.

2004-01-01

Layer-by-layer laser machining with mechanical removal of vitrified dross between passes is a new technique with a demonstrated capability for deep-section cutting, not only of concrete, but of ceramic and refractory materials in general. For this application fairly low power densities are required. A comparison of experimental results using high-power CO 2 and diode lasers under roughly equivalent experimental conditions, cutting to depths of >100 mm, is presented. A marked improvement in cutting depth per pass is observed for the case of the diode laser. The increased cutting rate is rationalized in terms of the combined effects of coupling efficiency and beam shape

GFDL's ESM2 global coupled climate-carbon Earth System Models. Part I: physical formulation and baseline simulation characteristics

Science.gov (United States)

Dunne, John P.; John, Jasmin G.; Adcroft, Alistair J.; Griffies, Stephen M.; Hallberg, Robert W.; Shevalikova, Elena; Stouffer, Ronald J.; Cooke, William; Dunne, Krista A.; Harrison, Matthew J.; Krasting, John P.; Malyshev, Sergey L.; Milly, P.C.D.; Phillipps, Peter J.; Sentman, Lori A.; Samuels, Bonita L.; Spelman, Michael J.; Winton, Michael; Wittenberg, Andrew T.; Zadeh, Niki

2012-01-01

We describe the physical climate formulation and simulation characteristics of two new global coupled carbon-climate Earth System Models, ESM2M and ESM2G. These models demonstrate similar climate fidelity as the Geophysical Fluid Dynamics Laboratory's previous CM2.1 climate model while incorporating explicit and consistent carbon dynamics. The two models differ exclusively in the physical ocean component; ESM2M uses Modular Ocean Model version 4.1 with vertical pressure layers while ESM2G uses Generalized Ocean Layer Dynamics with a bulk mixed layer and interior isopycnal layers. Differences in the ocean mean state include the thermocline depth being relatively deep in ESM2M and relatively shallow in ESM2G compared to observations. The crucial role of ocean dynamics on climate variability is highlighted in the El Niño-Southern Oscillation being overly strong in ESM2M and overly weak ESM2G relative to observations. Thus, while ESM2G might better represent climate changes relating to: total heat content variability given its lack of long term drift, gyre circulation and ventilation in the North Pacific, tropical Atlantic and Indian Oceans, and depth structure in the overturning and abyssal flows, ESM2M might better represent climate changes relating to: surface circulation given its superior surface temperature, salinity and height patterns, tropical Pacific circulation and variability, and Southern Ocean dynamics. Our overall assessment is that neither model is fundamentally superior to the other, and that both models achieve sufficient fidelity to allow meaningful climate and earth system modeling applications. This affords us the ability to assess the role of ocean configuration on earth system interactions in the context of two state-of-the-art coupled carbon-climate models.

Investigation of deep level defects in epitaxial semiconducting zinc sulpho-selenide. Progress report, 15 June 1979-14 June 1980

International Nuclear Information System (INIS)

Wessels, B.W.

1980-01-01

In an effort to understand the defect structure of the ternary II-VI compound zinc sulpho-selenide, the binary compound zinc selenide was investigated. Thin single crystalline films of zinc selenide were heteroepitaxially grown on (100) GaAs. Epitaxial layers from 5 to 50 microns thick could be readily grown using a chemical vapor transport technique. The layers had an excellent morphology with few stacking faults and hillocks. Detailed epitaxial growth kinetics were examined as a function of temperature and reactant concentration. It was found that hydrogen flow rate, source and substrate temperature affect the growth rate of the epitaxial films. Au - ZnSe Schottky barrier diodes and ZnSe - GaAs n-p heterojunctions were prepared from the epitaxial layers. Current-voltage characteristics were measured on both types of diodes. From capacitance-voltage measurements the residual doping density of the epitaxial layers were found to be of the order of 10 14 - 10 15 cm -3 . Finally, we have begun to measure the deep level spectrum of both the Schottky barrier diodes and the heterojunctions. Deep level transient spectroscopy appears to be well suited for determining trapping states in ZnSe provided the material has a low enough resistivity

Evidence for the bioerosion of deep-water corals by echinoids in the Northeast Atlantic

Science.gov (United States)

Stevenson, Angela; Rocha, Carlos

2013-01-01

In situ video observations of echinoids interacting with deep-sea coral are common in the deep-sea, but paradoxically the deep-sea literature is devoid of reports of bioerosion by extant echinoids. Here we present evidence of contemporary bioerosion of cold-water coral by four species of deep-sea echinoids, Gracilechinus elegans, Gracilechinus alexandri, Cidaris cidaris, and Araeosoma fenestratum, showing that they actively predate on the living framework of reef building corals, Lophelia pertusa and Madrepora oculata, in the NE Atlantic. Echinoid specimens were collected in six canyons located in the Bay of Biscay, France and two canyons on the north side of the Porcupine Bank and Goban Spur, Ireland. A total of 44 live specimens from the four taxa (9 of G. elegans, 4 of G. alexandri, 21 of C. cidaris and 10 of A. fenestratum) showed recent ingestion of the coral infrastructure. Upon dissection, live coral skeleton was observed encased in a thick mucus layer within the gastrointestinal tract of G. elegans and G. alexandri while both live and dead coral fragments were found in C. cidaris and A. fenestratum. Echinoid bioerosion limits the growth of shallow-water reefs. Our observations suggest that echinoids may also play an important role in the ecology of deep-water coral reefs.

Matrix completion by deep matrix factorization.

Science.gov (United States)

Fan, Jicong; Cheng, Jieyu

2018-02-01

Conventional methods of matrix completion are linear methods that are not effective in handling data of nonlinear structures. Recently a few researchers attempted to incorporate nonlinear techniques into matrix completion but there still exists considerable limitations. In this paper, a novel method called deep matrix factorization (DMF) is proposed for nonlinear matrix completion. Different from conventional matrix completion methods that are based on linear latent variable models, DMF is on the basis of a nonlinear latent variable model. DMF is formulated as a deep-structure neural network, in which the inputs are the low-dimensional unknown latent variables and the outputs are the partially observed variables. In DMF, the inputs and the parameters of the multilayer neural network are simultaneously optimized to minimize the reconstruction errors for the observed entries. Then the missing entries can be readily recovered by propagating the latent variables to the output layer. DMF is compared with state-of-the-art methods of linear and nonlinear matrix completion in the tasks of toy matrix completion, image inpainting and collaborative filtering. The experimental results verify that DMF is able to provide higher matrix completion accuracy than existing methods do and DMF is applicable to large matrices. Copyright © 2017 Elsevier Ltd. All rights reserved.

The effects of deep convection on the concentration and size distribution of aerosol particles within the upper troposphere: A case study

Science.gov (United States)

Yin, Yan; Chen, Qian; Jin, Lianji; Chen, Baojun; Zhu, Shichao; Zhang, Xiaopei

2012-11-01

A cloud resolving model coupled with a spectral bin microphysical scheme was used to investigate the effects of deep convection on the concentration and size distribution of aerosol particles within the upper troposphere. A deep convective storm that occurred on 1 December, 2005 in Darwin, Australia was simulated, and was compared with available radar observations. The results showed that the radar echo of the storm in the developing stage was well reproduced by the model. Sensitivity tests for aerosol layers at different altitudes were conducted in order to understand how the concentration and size distribution of aerosol particles within the upper troposphere can be influenced by the vertical transport of aerosols as a result of deep convection. The results indicated that aerosols originating from the boundary layer can be more efficiently transported upward, as compared to those from the mid-troposphere, due to significantly increased vertical velocity through the reinforced homogeneous freezing of droplets. Precipitation increased when aerosol layers were lofted at different altitudes, except for the case where an aerosol layer appeared at 5.4-8.0 km, in which relatively more efficient heterogeneous ice nucleation and subsequent Wegener-Bergeron-Findeisen process resulted in more pronounced production of ice crystals, and prohibited the formation of graupel particles via accretion. Sensitivity tests revealed, at least for the cases considered, that the concentration of aerosol particles within the upper troposphere increased by a factor of 7.71, 5.36, and 5.16, respectively, when enhanced aerosol layers existed at 0-2.2 km, 2.2-5.4 km, and 5.4-8.0 km, with Aitken mode and a portion of accumulation mode (0.1-0.2μm) particles being the most susceptible to upward transport.

BIG1 is required for the survival of deep layer neurons, neuronal polarity, and the formation of axonal tracts between the thalamus and neocortex in developing brain.

Directory of Open Access Journals (Sweden)

Jia-Jie Teoh

Full Text Available BIG1, an activator protein of the small GTPase, Arf, and encoded by the Arfgef1 gene, is one of candidate genes for epileptic encephalopathy. To know the involvement of BIG1 in epileptic encephalopathy, we analyzed BIG1-deficient mice and found that BIG1 regulates neurite outgrowth and brain development in vitro and in vivo. The loss of BIG1 decreased the size of the neocortex and hippocampus. In BIG1-deficient mice, the neuronal progenitor cells (NPCs and the interneurons were unaffected. However, Tbr1+ and Ctip2+ deep layer (DL neurons showed spatial-temporal dependent apoptosis. This apoptosis gradually progressed from the piriform cortex (PIR, peaked in the neocortex, and then progressed into the hippocampus from embryonic day 13.5 (E13.5 to E17.5. The upper layer (UL and DL order in the neocortex was maintained in BIG1-deficient mice, but the excitatory neurons tended to accumulate before their destination layers. Further pulse-chase migration assay showed that the migration defect was non-cell autonomous and secondary to the progression of apoptosis into the BIG1-deficient neocortex after E15.5. In BIG1-deficient mice, we observed an ectopic projection of corticothalamic axons from the primary somatosensory cortex (S1 into the dorsal lateral geniculate nucleus (dLGN. The thalamocortical axons were unable to cross the diencephalon-telencephalon boundary (DTB. In vitro, BIG1-deficient neurons showed a delay in neuronal polarization. BIG1-deficient neurons were also hypersensitive to low dose glutamate (5 μM, and died via apoptosis. This study showed the role of BIG1 in the survival of DL neurons in developing embryonic brain and in the generation of neuronal polarity.

Features of Red Sea Water Masses

Science.gov (United States)

Kartadikaria, Aditya; Hoteit, Ibrahim

2015-04-01

Features of Red Sea water mass can be divided into three types but best to be grouped into two different classes that are split at the potential density line σθ=27.4. The surface water (0-50 m) and the intermediate water (50-200 m) have nearly identical types of water mass. They appear as a maxima salinity layer for the water mass that has σθ > 26.0, and as a minimum salinity layer for water mass that has σθ water masses are strongly affected by mixing that is controlled by seasonal variability, fresh water intrusion of the Gulf of Aden Intermediate Water (GAIW), and eddies variability. Two types of mixing; isopycnal and diapycnal mixing are part of important physical phenomena that explain the change of water mass in the Red Sea. The isopycnal mixing occurs at the neutral potential density line, connecting the Red Sea with its adjacent channel, the Gulf of Aden. Diapycnal mixing is found as a dominant mixing mode in the surface of the Red Sea Water and mainly due to energetic eddy activity. Density gradients, across which diapycnal mixing occurs, in the Red Sea are mainly due to large variations in salinity. The isolation of an extreme haline water mass below the thermocline contributes to the generation of the latitudinal shift and low diapycnal mixing. This finding further explains the difference of spatial kinetic mixing between the RSW and the Indian Ocean basin.

Features of Red Sea Water Masses

KAUST Repository

Kartadikaria, Aditya R.

2015-04-01

Features of Red Sea water mass can be divided into three types but best to be grouped into two different classes that are split at the potential density line σθ=27.4. The surface water (0-50 m) and the intermediate water (50-200 m) have nearly identical types of water mass. They appear as a maxima salinity layer for the water mass that has σθ > 26.0, and as a minimum salinity layer for water mass that has σθ < 26.0. These types of water masses are strongly affected by mixing that is controlled by seasonal variability, fresh water intrusion of the Gulf of Aden Intermediate Water (GAIW), and eddies variability. Two types of mixing; isopycnal and diapycnal mixing are part of important physical phenomena that explain the change of water mass in the Red Sea. The isopycnal mixing occurs at the neutral potential density line, connecting the Red Sea with its adjacent channel, the Gulf of Aden. Diapycnal mixing is found as a dominant mixing mode in the surface of the Red Sea Water and mainly due to energetic eddy activity. Density gradients, across which diapycnal mixing occurs, in the Red Sea are mainly due to large variations in salinity. The isolation of an extreme haline water mass below the thermocline contributes to the generation of the latitudinal shift and low diapycnal mixing. This finding further explains the difference of spatial kinetic mixing between the RSW and the Indian Ocean basin.

EP-DNN: A Deep Neural Network-Based Global Enhancer Prediction Algorithm.

Science.gov (United States)

Kim, Seong Gon; Harwani, Mrudul; Grama, Ananth; Chaterji, Somali

2016-12-08

We present EP-DNN, a protocol for predicting enhancers based on chromatin features, in different cell types. Specifically, we use a deep neural network (DNN)-based architecture to extract enhancer signatures in a representative human embryonic stem cell type (H1) and a differentiated lung cell type (IMR90). We train EP-DNN using p300 binding sites, as enhancers, and TSS and random non-DHS sites, as non-enhancers. We perform same-cell and cross-cell predictions to quantify the validation rate and compare against two state-of-the-art methods, DEEP-ENCODE and RFECS. We find that EP-DNN has superior accuracy with a validation rate of 91.6%, relative to 85.3% for DEEP-ENCODE and 85.5% for RFECS, for a given number of enhancer predictions and also scales better for a larger number of enhancer predictions. Moreover, our H1 → IMR90 predictions turn out to be more accurate than IMR90 → IMR90, potentially because H1 exhibits a richer signature set and our EP-DNN model is expressive enough to extract these subtleties. Our work shows how to leverage the full expressivity of deep learning models, using multiple hidden layers, while avoiding overfitting on the training data. We also lay the foundation for exploration of cross-cell enhancer predictions, potentially reducing the need for expensive experimentation.

EP-DNN: A Deep Neural Network-Based Global Enhancer Prediction Algorithm

Science.gov (United States)

Kim, Seong Gon; Harwani, Mrudul; Grama, Ananth; Chaterji, Somali

2016-12-01

We present EP-DNN, a protocol for predicting enhancers based on chromatin features, in different cell types. Specifically, we use a deep neural network (DNN)-based architecture to extract enhancer signatures in a representative human embryonic stem cell type (H1) and a differentiated lung cell type (IMR90). We train EP-DNN using p300 binding sites, as enhancers, and TSS and random non-DHS sites, as non-enhancers. We perform same-cell and cross-cell predictions to quantify the validation rate and compare against two state-of-the-art methods, DEEP-ENCODE and RFECS. We find that EP-DNN has superior accuracy with a validation rate of 91.6%, relative to 85.3% for DEEP-ENCODE and 85.5% for RFECS, for a given number of enhancer predictions and also scales better for a larger number of enhancer predictions. Moreover, our H1 → IMR90 predictions turn out to be more accurate than IMR90 → IMR90, potentially because H1 exhibits a richer signature set and our EP-DNN model is expressive enough to extract these subtleties. Our work shows how to leverage the full expressivity of deep learning models, using multiple hidden layers, while avoiding overfitting on the training data. We also lay the foundation for exploration of cross-cell enhancer predictions, potentially reducing the need for expensive experimentation.

Deep Incremental Boosting

OpenAIRE

Mosca, Alan; Magoulas, George D

2017-01-01

This paper introduces Deep Incremental Boosting, a new technique derived from AdaBoost, specifically adapted to work with Deep Learning methods, that reduces the required training time and improves generalisation. We draw inspiration from Transfer of Learning approaches to reduce the start-up time to training each incremental Ensemble member. We show a set of experiments that outlines some preliminary results on some common Deep Learning datasets and discuss the potential improvements Deep In...

Crystallinity Improvement of Zn O Thin Film on Different Buffer Layers Grown by MBE

International Nuclear Information System (INIS)

Shao-Ying, T.; Che-Hao, L.; Wen-Ming, Ch.; Yang, C.C.; Po-Ju, Ch.; Hsiang-Chen, W.; Ya-Ping, H.

2012-01-01

The material and optical properties of Zn O thin film samples grown on different buffer layers on sapphire substrates through a two-step temperature variation growth by molecular beam epitaxy were investigated. The thin buffer layer between the Zn O layer and the sapphire substrate decreased the lattice mismatch to achieve higher quality Zn O thin film growth. A Ga N buffer layer slightly increased the quality of the Zn O thin film, but the threading dislocations still stretched along the c-axis of the Ga N layer. The use of Mg O as the buffer layer decreased the surface roughness of the Zn O thin film by 58.8% due to the suppression of surface cracks through strain transfer of the sample. From deep level emission and rocking curve measurements it was found that the threading dislocations play a more important role than oxygen vacancies for high-quality Zn O thin film growth.

Crystallinity Improvement of ZnO Thin Film on Different Buffer Layers Grown by MBE

Directory of Open Access Journals (Sweden)

Shao-Ying Ting

2012-01-01

Full Text Available The material and optical properties of ZnO thin film samples grown on different buffer layers on sapphire substrates through a two-step temperature variation growth by molecular beam epitaxy were investigated. The thin buffer layer between the ZnO layer and the sapphire substrate decreased the lattice mismatch to achieve higher quality ZnO thin film growth. A GaN buffer layer slightly increased the quality of the ZnO thin film, but the threading dislocations still stretched along the c-axis of the GaN layer. The use of MgO as the buffer layer decreased the surface roughness of the ZnO thin film by 58.8% due to the suppression of surface cracks through strain transfer of the sample. From deep level emission and rocking curve measurements it was found that the threading dislocations play a more important role than oxygen vacancies for high-quality ZnO thin film growth.

Formation of intrathermocline eddies at ocean fronts by wind-driven destruction of potential vorticity

Science.gov (United States)

Thomas, Leif N.

2008-08-01

A mechanism for the generation of intrathermocline eddies (ITEs) at wind-forced fronts is examined using a high resolution numerical simulation. Favorable conditions for ITE formation result at fronts forced by "down-front" winds, i.e. winds blowing in the direction of the frontal jet. Down-front winds exert frictional forces that reduce the potential vorticity (PV) within the surface boundary in the frontal outcrop, providing a source for the low-PV water that is the materia prima of ITEs. Meandering of the front drives vertical motions that subduct the low-PV water into the pycnocline, pooling it into the coherent anticyclonic vortex of a submesoscale ITE. As the fluid is subducted along the outcropping frontal isopycnal, the low-PV water, which at the surface is associated with strongly baroclinic flow, re-expresses itself as water with nearly zero absolute vorticity. This generation of strong anticyclonic vorticity results from the tilting of the horizontal vorticity of the frontal jet, not from vortex squashing. During the formation of the ITE, high-PV water from the pycnocline is upwelled alongside the subducting low-PV surface water. The positive correlation between the ITE's velocity and PV fields results in an upward, along-isopycnal eddy PV flux that scales with the surface frictional PV flux driven by the wind. The relationship between the eddy and wind-induced frictional PV flux is nonlocal in time, as the eddy PV flux persists long after the wind forcing is shut off. The ITE's PV flux affects the large-scale flow by driving an eddy-induced transport or bolus velocity down the outcropping isopycnal layer with a magnitude that scales with the Ekman velocity.

Validation of formability of laminated sheet metal for deep drawing process using GTN damage model

Energy Technology Data Exchange (ETDEWEB)

Lim, Yongbin; Cha, Wan-gi; Kim, Naksoo [Department of Mechanical Engineering, Sogang University, 1 Sinsu-dong, Mapo-gu, Seoul, 121-742 (Korea, Republic of); Ko, Sangjin [Mold/die and forming technology team, Product prestige research lab, LG electronics, 222, LG-ro, Jinwi-myeon, Pyeongtaek-si, Gyeonggi-do, 451-713 (Korea, Republic of)

2013-12-16

In this study, we studied formability of PET/PVC laminated sheet metal which named VCM (Vinyl Coated Metal). VCM offers various patterns and good-looking metal steel used for appliances such as refrigerator and washing machine. But, this sheet has problems which are crack and peeling of film when the material is formed by deep drawing process. To predict the problems, we used finite element method and GTN (Gurson-Tvergaard-Needleman) damage model to represent damage of material. We divided the VCM into 3 layers (PET film, adhesive and steel added PVC) in finite element analysis model to express the crack and peeling phenomenon. The material properties of each layer are determined by reverse engineering based on tensile test result. Furthermore, we performed the simple rectangular deep drawing and simulated it. The simulation result shows good agreement with drawing experiment result in position, punch stroke of crack occurrence. Also, we studied the fracture mechanism of PET film on VCM by comparing the width direction strain of metal and PET film.

Efficient fluorescent deep-blue and hybrid white emitting devices based on carbazole/benzimidazole compound

KAUST Repository

Yang, Xiaohui

2011-07-28

We report the synthesis, photophysics, and electrochemical characterization of carbazole/benzimidazole-based compound (Cz-2pbb) and efficient fluorescent deep-blue light emitting devices based on Cz-2pbb with the peak external quantum efficiency of 4.1% and Commission Internationale dÉnclairage coordinates of (0.16, 0.05). Efficient deep-blue emission as well as high triplet state energy of Cz-2pbb enables fabrication of hybrid white organic light emitting diodes with a single emissive layer. Hybrid white emitting devices based on Cz-2pbb show the peak external quantum efficiency exceeding 10% and power efficiency of 14.8 lm/W at a luminance of 500 cd/m2. © 2011 American Chemical Society.

Compression of deep convolutional neural network for computer-aided diagnosis of masses in digital breast tomosynthesis

Science.gov (United States)

Samala, Ravi K.; Chan, Heang-Ping; Hadjiiski, Lubomir; Helvie, Mark A.; Richter, Caleb; Cha, Kenny

2018-02-01

Deep-learning models are highly parameterized, causing difficulty in inference and transfer learning. We propose a layered pathway evolution method to compress a deep convolutional neural network (DCNN) for classification of masses in DBT while maintaining the classification accuracy. Two-stage transfer learning was used to adapt the ImageNet-trained DCNN to mammography and then to DBT. In the first-stage transfer learning, transfer learning from ImageNet trained DCNN was performed using mammography data. In the second-stage transfer learning, the mammography-trained DCNN was trained on the DBT data using feature extraction from fully connected layer, recursive feature elimination and random forest classification. The layered pathway evolution encapsulates the feature extraction to the classification stages to compress the DCNN. Genetic algorithm was used in an iterative approach with tournament selection driven by count-preserving crossover and mutation to identify the necessary nodes in each convolution layer while eliminating the redundant nodes. The DCNN was reduced by 99% in the number of parameters and 95% in mathematical operations in the convolutional layers. The lesion-based area under the receiver operating characteristic curve on an independent DBT test set from the original and the compressed network resulted in 0.88+/-0.05 and 0.90+/-0.04, respectively. The difference did not reach statistical significance. We demonstrated a DCNN compression approach without additional fine-tuning or loss of performance for classification of masses in DBT. The approach can be extended to other DCNNs and transfer learning tasks. An ensemble of these smaller and focused DCNNs has the potential to be used in multi-target transfer learning.

DeepRT: deep learning for peptide retention time prediction in proteomics

OpenAIRE

Ma, Chunwei; Zhu, Zhiyong; Ye, Jun; Yang, Jiarui; Pei, Jianguo; Xu, Shaohang; Zhou, Ruo; Yu, Chang; Mo, Fan; Wen, Bo; Liu, Siqi

2017-01-01

Accurate predictions of peptide retention times (RT) in liquid chromatography have many applications in mass spectrometry-based proteomics. Herein, we present DeepRT, a deep learning based software for peptide retention time prediction. DeepRT automatically learns features directly from the peptide sequences using the deep convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) model, which eliminates the need to use hand-crafted features or rules. After the feature learning, pr...

Recent progress in the development of 3D deep n-well CMOS MAPS

International Nuclear Information System (INIS)

Traversi, G; Manghisoni, M; Re, V; Gaioni, L; Manazza, A; Ratti, L; Zucca, S

2012-01-01

In the deep n-well (DNW) monolithic active pixel sensor (MAPS) a full in-pixel signal processing chain is integrated by exploiting the triple well option of a deep submicron CMOS process. This work is concerned with the design and characterization of DNW MAPS fabricated in a vertical integration (3D) CMOS technology. 3D processes can be very effective in overcoming typical limitations of monolithic active pixel sensors. This paper discusses the main features of a new analog processor for DNW MAPS (ApselVI) in view of applications to the SVT Layer0 of the SuperB Factory. It also presents the first experimental results from the test of a DNW MAPS prototype in the GlobalFoundries 130 nm CMOS technology.

Intelligent fault diagnosis of rolling bearings using an improved deep recurrent neural network

Science.gov (United States)

Jiang, Hongkai; Li, Xingqiu; Shao, Haidong; Zhao, Ke

2018-06-01

Traditional intelligent fault diagnosis methods for rolling bearings heavily depend on manual feature extraction and feature selection. For this purpose, an intelligent deep learning method, named the improved deep recurrent neural network (DRNN), is proposed in this paper. Firstly, frequency spectrum sequences are used as inputs to reduce the input size and ensure good robustness. Secondly, DRNN is constructed by the stacks of the recurrent hidden layer to automatically extract the features from the input spectrum sequences. Thirdly, an adaptive learning rate is adopted to improve the training performance of the constructed DRNN. The proposed method is verified with experimental rolling bearing data, and the results confirm that the proposed method is more effective than traditional intelligent fault diagnosis methods.

Hydrogeochemistry of deep groundwaters of mafic and ultramafic rocks in Finland

International Nuclear Information System (INIS)

Ruskeeniemi, T.; Blomqvist, R.; Lindberg, A.; Ahonen, L.; Frape, S.

1996-12-01

The present work reports and interprets the hydrogeochemical and hydrogeological data obtained from deep groundwaters in various mafic-ultramafic formations in Finland. The work is mainly based on the results of the research project 'Geochemistry of deep groundwaters' financed by the Ministry of Trade and Industry and the Geological Survey of Finland. Five sites were selected for this study: (1) Juuka, (2) Keminmaa, (3) Maentsaelae, (4) Ranua, and (5) Ylivieska. Keminmaa and Ranua are located in Early Proterozoic layered intrusions dated at 2.44 Ga. The Juuka site lies within the massive Miihkali serpentinite, which is thought to represent the ultramafic part of a Proterozoic (1.97 Ga) ophiolite complex. The Maentsaelae gabbro represents the deep parts of the Svecofennian volcanic sequence, while the Ylivieska mafic-ultramafic intrusion is one of a group of Svecokarelian Ni-potential intrusions 1.9 Ga in age. For reference, groundwaters from four other sites are also briefly described. Three of these sites are located within the nickel mining regions of Enonkoski, Kotalahti and Vammala, while the fourth is a small Ni mineralization at Hyvelae, Noormarkku. The four reference sites are all of Svecokarelian age. (refs.)

Neural network analysis of head-flow curves in deep well pumps

International Nuclear Information System (INIS)

Goelcue, Mustafa

2006-01-01

In impellers with splitter blades, the difficulty in calculation of the flow area of the impeller is because of the unknown flow rate occurring in the two separate areas when the splitter blades are added. Experimental studies were made to investigate the effects of splitter blade length on deep well pump performance for different numbers of blades. Head-flow curves of deep well pump impellers with splitter blades were investigated using artificial neural networks (ANNs). Gradient descent (GD), Gradient descent with momentum (GDM) and Levenberg-Marquardt (LM) learning algorithms were used in the networks. Experimental studies were completed to obtain training and test data. Blade number (z), non-dimensional splitter blade length (L-bar ) and flow rate (Q) were used as the input layer, while the output is head (H m ). For the testing data, the root mean squared error (RMSE), fraction of variance (R 2 ) and mean absolute percentage error (MAPE) were found to be 0.1285, 0.9999 and 1.6821%, respectively. With these results, we believe that the ANN can be used for prediction of head-flow curves as an appropriate method in deep well pump impellers with splitter blades.

Hydrogeochemistry of deep groundwaters of mafic and ultramafic rocks in Finland

Energy Technology Data Exchange (ETDEWEB)

Ruskeeniemi, T.; Blomqvist, R.; Lindberg, A.; Ahonen, L. [Geological Survey of Finland, Espoo (Finland); Frape, S. [Waterloo Univ., ON (Canada)

1996-12-01

The present work reports and interprets the hydrogeochemical and hydrogeological data obtained from deep groundwaters in various mafic-ultramafic formations in Finland. The work is mainly based on the results of the research project `Geochemistry of deep groundwaters` financed by the Ministry of Trade and Industry and the Geological Survey of Finland. Five sites were selected for this study: (1) Juuka, (2) Keminmaa, (3) Maentsaelae, (4) Ranua, and (5) Ylivieska. Keminmaa and Ranua are located in Early Proterozoic layered intrusions dated at 2.44 Ga. The Juuka site lies within the massive Miihkali serpentinite, which is thought to represent the ultramafic part of a Proterozoic (1.97 Ga) ophiolite complex. The Maentsaelae gabbro represents the deep parts of the Svecofennian volcanic sequence, while the Ylivieska mafic-ultramafic intrusion is one of a group of Svecokarelian Ni-potential intrusions 1.9 Ga in age. For reference, groundwaters from four other sites are also briefly described. Three of these sites are located within the nickel mining regions of Enonkoski, Kotalahti and Vammala, while the fourth is a small Ni mineralization at Hyvelae, Noormarkku. The four reference sites are all of Svecokarelian age. (refs.).

Complex approach mechanical properties and formability assessment of selected deep-drawing steels

OpenAIRE

J. Štaba; M. Buršák

2009-01-01

The paper analyses the properties of deep-drawing sheets of three grades (Re = 320 to 475 MPa), surface-treated with hot-dip galvanizing, made of microalloyed steels. Deformation properties are assessed using tensile tests, technological Erichsen or cupping tests. These characteristics, as well as the behaviour of the surface layer, are also investigated under dynamic conditions (modified Erichsen test using a drop tester), or using flat bending fatigue tests. Using microscopic analysis the d...

Characterization of deep wet etching of fused silica glass for single cell and optical sensor deposition

International Nuclear Information System (INIS)

Zhu, Haixin; Holl, Mark; Ray, Tathagata; Bhushan, Shivani; Meldrum, Deirdre R

2009-01-01

The development of a high-throughput single-cell metabolic rate monitoring system relies on the use of transparent substrate material for a single cell-trapping platform. The high optical transparency, high chemical resistance, improved surface quality and compatibility with the silicon micromachining process of fused silica make it very attractive and desirable for this application. In this paper, we report the results from the development and characterization of a hydrofluoric acid (HF) based deep wet-etch process on fused silica. The pin holes and notching defects of various single-coated masking layers during the etching are characterized and the most suitable masking materials are identified for different etch depths. The dependence of the average etch rate and surface roughness on the etch depth, impurity concentration and HF composition are also examined. The resulting undercut from the deep HF etch using various masking materials is also investigated. The developed and characterized process techniques have been successfully implemented in the fabrication of micro-well arrays for single cell trapping and sensor deposition. Up to 60 µm deep micro-wells have been etched in a fused silica substrate with over 90% process yield and repeatability. To our knowledge, such etch depth has never been achieved in a fused silica substrate by using a non-diluted HF etchant and a single-coated masking layer at room temperature

AlGaN-Based Deep-Ultraviolet Light Emitting Diodes Fabricated on AlN/sapphire Template

International Nuclear Information System (INIS)

Li-Wen, Sang; Zhi-Xin, Qin; Hao, Fang; Yan-Zhao, Zhang; Tao, Li; Zheng-Yu, Xu; Zhi-Jian, Yang; Bo, Shen; Guo-Yi, Zhang; Shu-Ping, Li; Wei-Huang, Yang; Hang-Yang, Chen; Da-Yi, Liu; Jun-Yong, Kang

2009-01-01

We report on the growth and fabrication of deep ultraviolet (DUV) light emitting diodes (LEDs) on an AlN template which was grown on a pulsed atomic-layer epitaxial buffer layer. Threading dislocation densities in the AlN layer are greatly decreased with the introduction of this buffer layer. The crystalline quality of the AlGaN epilayer is further improved by using a low-temperature GaN interlayer between AlGaN and AlN. Electroluminescences of different DUV-LED devices at a wavelength of between 262 and 317 nm are demonstrated. To improve the hole concentration of p-type AlGaN, Mg-doping with trimethylindium assistance approach is performed. It is found that the serial resistance of DUV-LED decreases and the performance of DUV-LED such as EL properties is improved. (condensed matter: electronic structure, electrical, magnetic, and optical properties)

The Next Era: Deep Learning in Pharmaceutical Research.

Science.gov (United States)

Ekins, Sean

2016-11-01

Over the past decade we have witnessed the increasing sophistication of machine learning algorithms applied in daily use from internet searches, voice recognition, social network software to machine vision software in cameras, phones, robots and self-driving cars. Pharmaceutical research has also seen its fair share of machine learning developments. For example, applying such methods to mine the growing datasets that are created in drug discovery not only enables us to learn from the past but to predict a molecule's properties and behavior in future. The latest machine learning algorithm garnering significant attention is deep learning, which is an artificial neural network with multiple hidden layers. Publications over the last 3 years suggest that this algorithm may have advantages over previous machine learning methods and offer a slight but discernable edge in predictive performance. The time has come for a balanced review of this technique but also to apply machine learning methods such as deep learning across a wider array of endpoints relevant to pharmaceutical research for which the datasets are growing such as physicochemical property prediction, formulation prediction, absorption, distribution, metabolism, excretion and toxicity (ADME/Tox), target prediction and skin permeation, etc. We also show that there are many potential applications of deep learning beyond cheminformatics. It will be important to perform prospective testing (which has been carried out rarely to date) in order to convince skeptics that there will be benefits from investing in this technique.

Sedimentologic and volcanologic investigation of the deep tyrrhenian sea: preliminary result of cruise VST02

Directory of Open Access Journals (Sweden)

A. Bertagnini

2006-06-01

Full Text Available The VST02 cruise carried out in the summer of 2002 was focused at sedimentologic and volcanologic researches over selected areas of the deep portion of the Tyrrhenian sea. Chirp lines and seafloor samples were collected from the Calabrian slope surrounding Stromboli island, in the Marsili deep sea fan, in the Vavilov basin and in the Vavilov seamount. Submarine volcanic activity, both explosive and effusive, is occuring in the Stromboli edifice. Explosive submarine volcanism affects also the shallowest areas of the Vavilov seamount. Submarine carbonate lithification has been observed on the sediment-starved flanks of the Vavilov seamount. Acoustic transparent layers make up the recentmost infill of the Gortani basin, the easternmost portion of the Vavilov basin. Channels comprised of a variety of architectural elements and depositional lobes are the main elements of the Marsili deep-sea fan where, apparently, sedimentation occurs mainly through debris flow processes.

Vertical migrations of a deep-sea fish and its prey.

Directory of Open Access Journals (Sweden)

Pedro Afonso

Full Text Available It has been speculated that some deep-sea fishes can display large vertical migrations and likely doing so to explore the full suite of benthopelagic food resources, especially the pelagic organisms of the deep scattering layer (DSL. This would help explain the success of fishes residing at seamounts and the increased biodiversity found in these features of the open ocean. We combined active plus passive acoustic telemetry of blackspot seabream with in situ environmental and biological (backscattering data collection at a seamount to verify if its behaviour is dominated by vertical movements as a response to temporal changes in environmental conditions and pelagic prey availability. We found that seabream extensively migrate up and down the water column, that these patterns are cyclic both in short-term (tidal, diel as well as long-term (seasonal scales, and that they partially match the availability of potential DSL prey components. Furthermore, the emerging pattern points to a more complex spatial behaviour than previously anticipated, suggesting a seasonal switch in the diel behaviour mode (benthic vs. pelagic of seabream, which may reflect an adaptation to differences in prey availability. This study is the first to document the fine scale three-dimensional behaviour of a deep-sea fish residing at seamounts.

Characterization of deep level defects and thermally stimulated depolarization phenomena in La-doped TlInS{sub 2} layered semiconductor

Energy Technology Data Exchange (ETDEWEB)

Seyidov, MirHasan Yu., E-mail: smirhasan@gyte.edu.tr; Suleymanov, Rauf A.; Mikailzade, Faik A. [Department of Physics, Gebze Technical University, Gebze, Kocaeli 41400 (Turkey); Institute of Physics of NAS of Azerbaijan, H. Javid ave. 33, Baku AZ-1143 (Azerbaijan); Kargın, Elif Orhan [Department of Physics, Gebze Technical University, Gebze, Kocaeli 41400 (Turkey); Odrinsky, Andrei P. [Institute of Technical Acoustics, National Academy of Sciences of Belarus, Lyudnikov ave. 13, Vitebsk 210717 (Belarus)

2015-06-14

Lanthanum-doped high quality TlInS{sub 2} (TlInS{sub 2}:La) ferroelectric-semiconductor was characterized by photo-induced current transient spectroscopy (PICTS). Different impurity centers are resolved and identified. Analyses of the experimental data were performed in order to determine the characteristic parameters of the extrinsic and intrinsic defects. The energies and capturing cross section of deep traps were obtained by using the heating rate method. The observed changes in the Thermally Stimulated Depolarization Currents (TSDC) near the phase transition points in TlInS{sub 2}:La ferroelectric-semiconductor are interpreted as a result of self-polarization of the crystal due to the internal electric field caused by charged defects. The TSDC spectra show the depolarization peaks, which are attributed to defects of dipolar origin. These peaks provide important information on the defect structure and localized energy states in TlInS{sub 2}:La. Thermal treatments of TlInS{sub 2}:La under an external electric field, which was applied at different temperatures, allowed us to identify a peak in TSDC which was originated from La-dopant. It was established that deep energy level trap BTE43, which are active at low temperature (T ≤ 156 K) and have activation energy 0.29 eV and the capture cross section 2.2 × 10{sup −14} cm{sup 2}, corresponds to the La dopant. According to the PICTS results, the deep level trap center B5 is activated in the temperature region of incommensurate (IC) phases of TlInS{sub 2}:La, having the giant static dielectric constant due to the structural disorders. From the PICTS simulation results for B5, native deep level trap having an activation energy of 0.3 eV and the capture cross section of 1.8 × 10{sup −16} cm{sup 2} were established. A substantial amount of residual space charges is trapped by the deep level localized energy states of B5 in IC-phase. While the external electric field is applied, permanent dipoles

Analyses of the deep borehole drilling status for a deep borehole disposal system

Energy Technology Data Exchange (ETDEWEB)

Lee, Jong Youl; Choi, Heui Joo; Lee, Min Soo; Kim, Geon Young; Kim, Kyung Su [KAERI, Daejeon (Korea, Republic of)

2016-05-15

The purpose of disposal for radioactive wastes is not only to isolate them from humans, but also to inhibit leakage of any radioactive materials into the accessible environment. Because of the extremely high level and long-time scale radioactivity of HLW(High-level radioactive waste), a mined deep geological disposal concept, the disposal depth is about 500 m below ground, is considered as the safest method to isolate the spent fuels or high-level radioactive waste from the human environment with the best available technology at present time. Therefore, as an alternative disposal concept, i.e., deep borehole disposal technology is under consideration in number of countries in terms of its outstanding safety and cost effectiveness. In this paper, the general status of deep drilling technologies was reviewed for deep borehole disposal of high level radioactive wastes. Based on the results of these review, very preliminary applicability of deep drilling technology for deep borehole disposal analyzed. In this paper, as one of key technologies of deep borehole disposal system, the general status of deep drilling technologies in oil industry, geothermal industry and geo scientific field was reviewed for deep borehole disposal of high level radioactive wastes. Based on the results of these review, the very preliminary applicability of deep drilling technology for deep borehole disposal such as relation between depth and diameter, drilling time and feasibility classification was analyzed.

On the Hole Injection for III-Nitride Based Deep Ultraviolet Light-Emitting Diodes.

Science.gov (United States)

Li, Luping; Zhang, Yonghui; Xu, Shu; Bi, Wengang; Zhang, Zi-Hui; Kuo, Hao-Chung

2017-10-24

The hole injection is one of the bottlenecks that strongly hinder the quantum efficiency and the optical power for deep ultraviolet light-emitting diodes (DUV LEDs) with the emission wavelength smaller than 360 nm. The hole injection efficiency for DUV LEDs is co-affected by the p-type ohmic contact, the p-type hole injection layer, the p-type electron blocking layer and the multiple quantum wells. In this report, we review a large diversity of advances that are currently adopted to increase the hole injection efficiency for DUV LEDs. Moreover, by disclosing the underlying device physics, the design strategies that we can follow have also been suggested to improve the hole injection for DUV LEDs.

Molecular Phylogeny Of Microbes In The Deep-Sea Sediments From Tropical West Pacific Warm Pool

Science.gov (United States)

Wang, F.; Xiao, X.; Wang, P.

2005-12-01

The presence and phylogeny of bacteria and archaea in five deep-sea sediment samples collected from west Pacific Warm Pool area (WP-0, WP-1, WP-2, WP-3, WP-4), and in five sediment layers (1cm-, 3cm-, 6cm-, 10cm-, 12cm- layer) of the 12-cm sediment core of WP-0 were checked and compared. The microbial diversity in the five deep-sea sediments were similar as revealed by denaturing gradient gel electrophoresis, and all of them contained members of non-thermophilic marine group I crenarchaeota as the predominant archaeal group. The composition of methylotrophs including methanotrophs, sulfate reducing bacteria in the WP-0 sediment core were further investigated by molecular marker based analysis of mxaF, pmoA, dsrAB, specific anoxic methane oxidation archaeal and sulfate reducing bacterial 16S rRNA genes. From MxaF amino acid sequence analysis, it was demonstrated that microbes belonging to α - Proteobacteria most related to Hyphomicrobium and Methylobacterium were dominant aerobic methylotrophs in this deep-sea sediment; and small percentage of type II methanotrophs affiliating closest to Methylocystis and Methylosinus were also detected in this environment. mxaF quantitative PCR results showed that in the west Pacific WP sediment there existed around 3× 10 4-5 methylotrophs per gram sediment, 10-100 times more than that in samples collected from several other deep-sea Pacific sediment sample, but about 10 times less than that present in samples collected from rice and flower garden soil. Diverse groups of novel archaea (named as WPA), not belonging to any known archaeal lineages were checked out. They could be placed in the euryarchaeota kingdom, separated into two distinct groups, the main group was peripherally related with methanogens, the other group related with Thermoplasma. Possible sulfate reducing bacterial related with Desulfotomaculum, Desulfacinum, Desulfomonile and Desulfanuticus were also detected in our study. The vertical distributions of WPA

Assumed Probability Density Functions for Shallow and Deep Convection

Directory of Open Access Journals (Sweden)

Steven K Krueger

2010-10-01

Full Text Available The assumed joint probability density function (PDF between vertical velocity and conserved temperature and total water scalars has been suggested to be a relatively computationally inexpensive and unified subgrid-scale (SGS parameterization for boundary layer clouds and turbulent moments. This paper analyzes the performance of five families of PDFs using large-eddy simulations of deep convection, shallow convection, and a transition from stratocumulus to trade wind cumulus. Three of the PDF families are based on the double Gaussian form and the remaining two are the single Gaussian and a Double Delta Function (analogous to a mass flux model. The assumed PDF method is tested for grid sizes as small as 0.4 km to as large as 204.8 km. In addition, studies are performed for PDF sensitivity to errors in the input moments and for how well the PDFs diagnose some higher-order moments. In general, the double Gaussian PDFs more accurately represent SGS cloud structure and turbulence moments in the boundary layer compared to the single Gaussian and Double Delta Function PDFs for the range of grid sizes tested. This is especially true for small SGS cloud fractions. While the most complex PDF, Lewellen-Yoh, better represents shallow convective cloud properties (cloud fraction and liquid water mixing ratio compared to the less complex Analytic Double Gaussian 1 PDF, there appears to be no advantage in implementing Lewellen-Yoh for deep convection. However, the Analytic Double Gaussian 1 PDF better represents the liquid water flux, is less sensitive to errors in the input moments, and diagnoses higher order moments more accurately. Between the Lewellen-Yoh and Analytic Double Gaussian 1 PDFs, it appears that neither family is distinctly better at representing cloudy layers. However, due to the reduced computational cost and fairly robust results, it appears that the Analytic Double Gaussian 1 PDF could be an ideal family for SGS cloud and turbulence

Deep Space Telecommunications

Science.gov (United States)

Kuiper, T. B. H.; Resch, G. M.

2000-01-01

The increasing load on NASA's deep Space Network, the new capabilities for deep space missions inherent in a next-generation radio telescope, and the potential of new telescope technology for reducing construction and operation costs suggest a natural marriage between radio astronomy and deep space telecommunications in developing advanced radio telescope concepts.

Electronic relaxation of deep bulk trap and interface state in ZnO ceramics

International Nuclear Information System (INIS)

Yang Yan; Li Sheng-Tao; Ding Can; Cheng Peng-Fei

2011-01-01

This paper investigates the electronic relaxation of deep bulk trap and interface state in ZnO ceramics based on dielectric spectra measured in a wide range of temperature, frequency and bias, in addition to the steady state response. It discusses the nature of net current flowing over the barrier affected by interface state, and then obtains temperature-dependent barrier height by approximate calculation from steady I—V (current—voltage) characteristics. Additional conductance and capacitance arising from deep bulk trap relaxation are calculated based on the displacement of the cross point between deep bulk trap and Fermi level under small AC signal. From the resonances due to deep bulk trap relaxation on dielectric spectra, the activation energies are obtained as 0.22 eV and 0.35 eV, which are consistent with the electronic levels of the main defect interstitial Zn and vacancy oxygen in the depletion layer. Under moderate bias, another resonance due to interface relaxation is shown on the dielectric spectra. The DC-like conductance is also observed in high temperature region on dielectric spectra, and the activation energy is much smaller than the barrier height in steady state condition, which is attributed to the displacement current coming from the shallow bulk trap relaxation or other factors. (fluids, plasmas and electric discharges)

Improved PECVD Si x N y film as a mask layer for deep wet etching of the silicon

Science.gov (United States)

Han, Jianqiang; Yin, Yi Jun; Han, Dong; Dong, LiZhen

2017-09-01

Although plasma enhanced chemical vapor deposition (PECVD) silicon nitride (Si x N y ) films have been extensively investigated by many researchers, requirements of film properties vary from device to device. For some applications utilizing Si x N y film as the mask Layer for deep wet etching of the silicon, it is very desirable to obtain a high quality film. In this study, Si x N y films were deposited on silicon substrates by PECVD technique from the mixtures of NH3 and 5% SiH4 diluted in Ar. The deposition temperature and RF power were fixed at 400 °C and 20 W, respectively. By adjusting the SiH4/NH3 flow ratio, Si x N y films of different compositions were deposited on silicon wafers. The stoichiometry, residual stress, etch rate in 1:50 HF, BHF solution and 40% KOH solution of deposited Si x N y films were measured. The experimental results show that the optimum SiH4/NH3 flow ratio at which deposited Si x N y films can perfectly protect the polysilicon resistors on the front side of wafers during KOH etching is between 1.63 and 2.24 under the given temperature and RF power. Polysilicon resistors protected by the Si x N y films can withstand 6 h 40% KOH double-side etching at 80 °C. At the range of SiH4/NH3 flow ratios, the Si/N atom ratio of films ranges from 0.645 to 0.702, which slightly deviate the ideal stoichiometric ratio of LPCVD Si3N4 film. In addition, the silicon nitride films with the best protection effect are not the films of minimum etch rate in KOH solution.

The effect of the electrode material on the electrodeposition of zinc from deep eutectic solvents

International Nuclear Information System (INIS)

Vieira, L.; Schennach, R.; Gollas, B.

2016-01-01

Highlights: • Mechanistic insight into zinc electrodeposition from deep eutectic solvents. • Overpotential for hydrogen evolution affects the electrodeposition of zinc. • Electrodeposited zinc forms surface alloys on Cu, Au, and Pt. • In situ PM-IRRAS of a ZnCl_2 containing deep eutectic solvent on glassy carbon. - Abstract: The voltammetric behaviour of the ZnCl_2 containing deep eutectic solvent choline chloride/ethylene glycol 1:2 was investigated on glassy carbon, stainless steel, Au, Pt, Cu, and Zn electrodes. While cyclic voltammetry on glassy carbon and stainless steel showed a cathodic peak for zinc electrodeposition only in the anodic reverse sweep, a cathodic peak was found also in the cathodic forward sweep on Au, Pt, Cu, and Zn. This behaviour is in agreement with the proposed mechanism of zinc deposition from an intermediate species Z, whose formation depends on the cathodic reduction potential of the solvent. The voltammetric reduction of the electrolyte involves hydrogen evolution and as a result the formation of Z and its reduction to zinc depend on the hydrogen overpotential for each electrode material. On Au, Pt, and Cu also the anodic stripping was different from that on glassy carbon and steel due to the formation of surface zinc alloys with the three former metals. The morphology of the zinc layers on Cu has been characterised by scanning electron microscopy and focussed ion beam. X-ray diffraction confirmed the presence of crystalline zinc and a Cu_4Zn phase. Spectroelectrochemistry by means of polarization modulation reflection-absorption spectroscopy (PM-IRRAS) on a glassy carbon electrode in the ZnCl_2 containing deep eutectic solvent showed characteristic potential dependent changes. The variation of band intensities at different applied potentials correlate with the voltammetry and suggest the formation of a compact blocking layer on the electrode surface, which inhibits the electrodeposition of zinc at sufficiently negative

Possibilities for the storage of radioactive waste in deep clay formations

International Nuclear Information System (INIS)

Le Pochat, G.; Lienhardt, M.J.; Peaudecerf, P.; Platel, J.P.; Simon, J.M.; Berest, P.; Charpentier, J.P.; Andre-Jehan, R.

1984-02-01

The possible storage sites in deep clay formations have been studied in parts of large French sedimentary basins which prima facie seem to have suitable characteristics. The most suitable areas were chosen on the basis of earlier oil prospecting data consisting of information on seismic movements, diagraphic well-logging data and old samples that happened to have been preserved. At the same time, the lithology of the clay formations can be determined from mineralogical studies on samples taken from boreholes or from outcrops. Before carrying out in situ experiments concerned with the geotechnical characterization of the deep clays, measurements were made in the laboratory on samples obtained in two ways: from tertiary clay outcrops and from cores taken at 950 m in the clay layers during oil well logging. The results of studies carried out on tertiary clays in Les Landes illustrate this procedure

The Great Wall: Urca Cooling Layers in the Accreted NS Crust

Directory of Open Access Journals (Sweden)

Meisel Zach

2018-01-01

Full Text Available Accreting neutron stars host a number of astronomical observables which can be used to infer the properties of the underlying dense matter. These observables are sensitive to the heating and cooling processes taking place in the accreted neutron star (NS crust. Within the past few years it has become apparent that electron-capture/beta-decay (urca cycles can operate within the NS crust at high temperatures. Layers of nuclei undergoing urca cycling can create a thermal barrier, or Great Wall, between heating occurring deep in the crust and the regions above the urca layers. This paper briefly reviews the urca process and the implications for observables from accreting neutron stars.

DeepBipolar: Identifying genomic mutations for bipolar disorder via deep learning.

Science.gov (United States)

Laksshman, Sundaram; Bhat, Rajendra Rana; Viswanath, Vivek; Li, Xiaolin

2017-09-01

Bipolar disorder, also known as manic depression, is a brain disorder that affects the brain structure of a patient. It results in extreme mood swings, severe states of depression, and overexcitement simultaneously. It is estimated that roughly 3% of the population of the United States (about 5.3 million adults) suffers from bipolar disorder. Recent research efforts like the Twin studies have demonstrated a high heritability factor for the disorder, making genomics a viable alternative for detecting and treating bipolar disorder, in addition to the conventional lengthy and costly postsymptom clinical diagnosis. Motivated by this study, leveraging several emerging deep learning algorithms, we design an end-to-end deep learning architecture (called DeepBipolar) to predict bipolar disorder based on limited genomic data. DeepBipolar adopts the Deep Convolutional Neural Network (DCNN) architecture that automatically extracts features from genotype information to predict the bipolar phenotype. We participated in the Critical Assessment of Genome Interpretation (CAGI) bipolar disorder challenge and DeepBipolar was considered the most successful by the independent assessor. In this work, we thoroughly evaluate the performance of DeepBipolar and analyze the type of signals we believe could have affected the classifier in distinguishing the case samples from the control set. © 2017 Wiley Periodicals, Inc.

Deep learning? What deep learning? | Fourie | South African ...

African Journals Online (AJOL)

In teaching generally over the past twenty years, there has been a move towards teaching methods that encourage deep, rather than surface approaches to learning. The reason for this being that students, who adopt a deep approach to learning are considered to have learning outcomes of a better quality and desirability ...

APPLICATION OF DEEP LEARNING IN GLOBELAND30-2010 PRODUCT REFINEMENT

Directory of Open Access Journals (Sweden)

T. Liu

2018-04-01

Full Text Available GlobeLand30, as one of the best Global Land Cover (GLC product at 30-m resolution, has been widely used in many research fields. Due to the significant spectral confusion among different land cover types and limited textual information of Landsat data, the overall accuracy of GlobeLand30 is about 80 %. Although such accuracy is much higher than most other global land cover products, it cannot satisfy various applications. There is still a great need of an effective method to improve the quality of GlobeLand30. The explosive high-resolution satellite images and remarkable performance of Deep Learning on image classification provide a new opportunity to refine GlobeLand30. However, the performance of deep leaning depends on quality and quantity of training samples as well as model training strategy. Therefore, this paper 1 proposed an automatic training sample generation method via Google earth to build a large training sample set; and 2 explore the best training strategy for land cover classification using GoogleNet (Inception V3, one of the most widely used deep learning network. The result shows that the fine-tuning from first layer of Inception V3 using rough large sample set is the best strategy. The retrained network was then applied in one selected area from Xi’an city as a case study of GlobeLand30 refinement. The experiment results indicate that the proposed approach with Deep Learning and google earth imagery is a promising solution for further improving accuracy of GlobeLand30.

Application of Deep Learning in GLOBELAND30-2010 Product Refinement

Science.gov (United States)

Liu, T.; Chen, X.

2018-04-01

GlobeLand30, as one of the best Global Land Cover (GLC) product at 30-m resolution, has been widely used in many research fields. Due to the significant spectral confusion among different land cover types and limited textual information of Landsat data, the overall accuracy of GlobeLand30 is about 80 %. Although such accuracy is much higher than most other global land cover products, it cannot satisfy various applications. There is still a great need of an effective method to improve the quality of GlobeLand30. The explosive high-resolution satellite images and remarkable performance of Deep Learning on image classification provide a new opportunity to refine GlobeLand30. However, the performance of deep leaning depends on quality and quantity of training samples as well as model training strategy. Therefore, this paper 1) proposed an automatic training sample generation method via Google earth to build a large training sample set; and 2) explore the best training strategy for land cover classification using GoogleNet (Inception V3), one of the most widely used deep learning network. The result shows that the fine-tuning from first layer of Inception V3 using rough large sample set is the best strategy. The retrained network was then applied in one selected area from Xi'an city as a case study of GlobeLand30 refinement. The experiment results indicate that the proposed approach with Deep Learning and google earth imagery is a promising solution for further improving accuracy of GlobeLand30.

Deep learning for biomarker regression: application to osteoporosis and emphysema on chest CT scans

Science.gov (United States)

González, Germán.; Washko, George R.; San José Estépar, Raúl

2018-03-01

Introduction: Biomarker computation using deep-learning often relies on a two-step process, where the deep learning algorithm segments the region of interest and then the biomarker is measured. We propose an alternative paradigm, where the biomarker is estimated directly using a regression network. We showcase this image-tobiomarker paradigm using two biomarkers: the estimation of bone mineral density (BMD) and the estimation of lung percentage of emphysema from CT scans. Materials and methods: We use a large database of 9,925 CT scans to train, validate and test the network for which reference standard BMD and percentage emphysema have been already computed. First, the 3D dataset is reduced to a set of canonical 2D slices where the organ of interest is visible (either spine for BMD or lungs for emphysema). This data reduction is performed using an automatic object detector. Second, The regression neural network is composed of three convolutional layers, followed by a fully connected and an output layer. The network is optimized using a momentum optimizer with an exponential decay rate, using the root mean squared error as cost function. Results: The Pearson correlation coefficients obtained against the reference standards are r = 0.940 (p < 0.00001) and r = 0.976 (p < 0.00001) for BMD and percentage emphysema respectively. Conclusions: The deep-learning regression architecture can learn biomarkers from images directly, without indicating the structures of interest. This approach simplifies the development of biomarker extraction algorithms. The proposed data reduction based on object detectors conveys enough information to compute the biomarkers of interest.

A study of ion implanted gallium arsenide using deep level transient spectroscopy

International Nuclear Information System (INIS)

Emerson, N.G.

1981-03-01

This thesis is concerned with the study of deep energy levels in ion implanted gallium arsenide (GaAs) using deep level transient spectroscopy (D.L.T.S.). The D.L.T.S. technique is used to characterise deep levels in terms of their activation energies and capture cross-sections and to determine their concentration profiles. The main objective is to characterise the effects on deep levels, of ion implantation and the related annealing processes. In the majority of cases assessment is carried out using Schottky barrier diodes. Low doses of selenium ions 1 to 3 x 10 12 cm -2 are implanted into vapour phase epitaxial (V.P.E.) GaAs and the effects of post-implantation thermal and pulsed laser annealing are compared. The process of oxygen implantation with doses in the range 1 x 10 12 to 5 x 10 13 cm -2 followed by thermal annealing at about 750 deg C, introduces a deep level at 0.79 eV from the conduction band. Oxygen implantation, at doses of 5 x 10 13 cm -2 , into V.P.E. GaAs produces a significant increase in the concentration of the A-centre (0.83 eV). High doses of zinc (10 15 cm -2 ) are implanted into n-type V.P.E. GaAs to form shallow p-type layers. The D.L.T.S. system described in the text is used to measure levels in the range 0.16 to 1.1 eV (for GaAs) with a sensitivity of the order 1:10 3 . (U.K.)

Silicon pool dynamics and biogenic silica export in the Southern Ocean inferred from Si-isotopes

Directory of Open Access Journals (Sweden)

F. Fripiat

2011-09-01

Full Text Available Silicon isotopic signatures (δ³⁰Si of water column silicic acid (Si(OH₄ were measured in the Southern Ocean, along a meridional transect from South Africa (Subtropical Zone down to 57° S (northern Weddell Gyre. This provides the first reported data of a summer transect across the whole Antarctic Circumpolar Current (ACC. δ³⁰Si variations are large in the upper 1000 m, reflecting the effect of the silica pump superimposed upon meridional water transfer across the ACC: the transport of Antarctic surface waters northward by a net Ekman drift and their convergence and mixing with warmer upper-ocean Si-depleted waters to the north. Using Si isotopic signatures, we determine different mixing interfaces: the Antarctic Surface Water (AASW, the Antarctic Intermediate Water (AAIW, and thermoclines in the low latitude areas. The residual silicic acid concentrations of end-members control the δ³⁰Si alteration of the mixing products and with the exception of AASW, all mixing interfaces have a highly Si-depleted mixed layer end-member. These processes deplete the silicic acid AASW concentration northward, across the different interfaces, without significantly changing the AASW δ³⁰Si composition. By comparing our new results with a previous study in the Australian sector we show that during the circumpolar transport of the ACC eastward, the δ³⁰Si composition of the silicic acid pools is getting slightly, but significantly lighter from the Atlantic to the Australian sectors. This results either from the dissolution of biogenic silica in the deeper layers and/or from an isopycnal mixing with the deep water masses in the different oceanic basins: North Atlantic Deep Water in the Atlantic, and Indian Ocean deep water in the Indo-Australian sector. This isotopic trend is further transmitted to the subsurface waters, representing mixing interfaces between the surface and deeper layers.

DeepInfer: open-source deep learning deployment toolkit for image-guided therapy

Science.gov (United States)

Mehrtash, Alireza; Pesteie, Mehran; Hetherington, Jorden; Behringer, Peter A.; Kapur, Tina; Wells, William M.; Rohling, Robert; Fedorov, Andriy; Abolmaesumi, Purang

2017-03-01

Deep learning models have outperformed some of the previous state-of-the-art approaches in medical image analysis. Instead of using hand-engineered features, deep models attempt to automatically extract hierarchical representations at multiple levels of abstraction from the data. Therefore, deep models are usually considered to be more flexible and robust solutions for image analysis problems compared to conventional computer vision models. They have demonstrated significant improvements in computer-aided diagnosis and automatic medical image analysis applied to such tasks as image segmentation, classification and registration. However, deploying deep learning models often has a steep learning curve and requires detailed knowledge of various software packages. Thus, many deep models have not been integrated into the clinical research work ows causing a gap between the state-of-the-art machine learning in medical applications and evaluation in clinical research procedures. In this paper, we propose "DeepInfer" - an open-source toolkit for developing and deploying deep learning models within the 3D Slicer medical image analysis platform. Utilizing a repository of task-specific models, DeepInfer allows clinical researchers and biomedical engineers to deploy a trained model selected from the public registry, and apply it to new data without the need for software development or configuration. As two practical use cases, we demonstrate the application of DeepInfer in prostate segmentation for targeted MRI-guided biopsy and identification of the target plane in 3D ultrasound for spinal injections.

Expansible apparatus for removing the surface layer from a concrete object

International Nuclear Information System (INIS)

Allen, C.H.

1979-01-01

A method and apparatus for removing the surface layer from a concrete object are described. The method consists of providing a hole having a circular wall in the surface layer of the object, the hole being at least as deep as the thickness of the surface layer to be removed, and applying an outward wedging pressure on the wall of the hole sufficient to spall the surface layer around the hole. By the proper spacing of an appropriate number of holes, it is possible to remove the entire surface layer. The apparatus consists of an elongated tubular-shaped body having a relatively short handle with a solid wall at one end. The wall of the remainder of the body contains a plurality of evenly spaced longitudinal cuts to form a relatively long expandable section. The outer end of the expandable section has an expandable, wedge-shaped spalling edge extending from the outer surface of the wall, perpendicular to the longitudinal axis of the body, and expanding means in the body for outwardly expanding the expandable section and forcing the spalling edge into the wall of a hole with sufficient outward pressure to spall away the surface layer of concrete. The method and apparatus are particularly suitable for removing surface layers of concrete which are radioactively contaminated

Deep Fault Recognizer: An Integrated Model to Denoise and Extract Features for Fault Diagnosis in Rotating Machinery

Directory of Open Access Journals (Sweden)

Xiaojie Guo

2016-12-01

Full Text Available Fault diagnosis in rotating machinery is significant to avoid serious accidents; thus, an accurate and timely diagnosis method is necessary. With the breakthrough in deep learning algorithm, some intelligent methods, such as deep belief network (DBN and deep convolution neural network (DCNN, have been developed with satisfactory performances to conduct machinery fault diagnosis. However, only a few of these methods consider properly dealing with noises that exist in practical situations and the denoising methods are in need of extensive professional experiences. Accordingly, rethinking the fault diagnosis method based on deep architectures is essential. Hence, this study proposes an automatic denoising and feature extraction method that inherently considers spatial and temporal correlations. In this study, an integrated deep fault recognizer model based on the stacked denoising autoencoder (SDAE is applied to both denoise random noises in the raw signals and represent fault features in fault pattern diagnosis for both bearing rolling fault and gearbox fault, and trained in a greedy layer-wise fashion. Finally, the experimental validation demonstrates that the proposed method has better diagnosis accuracy than DBN, particularly in the existing situation of noises with superiority of approximately 7% in fault diagnosis accuracy.

Understanding a Deep Learning Technique through a Neuromorphic System a Case Study with SpiNNaker Neuromorphic Platform

Directory of Open Access Journals (Sweden)

Sugiarto Indar

2018-01-01

Full Text Available Deep learning (DL has been considered as a breakthrough technique in the field of artificial intelligence and machine learning. Conceptually, it relies on a many-layer network that exhibits a hierarchically non-linear processing capability. Some DL architectures such as deep neural networks, deep belief networks and recurrent neural networks have been developed and applied to many fields with incredible results, even comparable to human intelligence. However, many researchers are still sceptical about its true capability: can the intelligence demonstrated by deep learning technique be applied for general tasks? This question motivates the emergence of another research discipline: neuromorphic computing (NC. In NC, researchers try to identify the most fundamental ingredients that construct intelligence behaviour produced by the brain itself. To achieve this, neuromorphic systems are developed to mimic the brain functionality down to cellular level. In this paper, a neuromorphic platform called SpiNNaker is described and evaluated in order to understand its potential use as a platform for a deep learning approach. This paper is a literature review that contains comparative study on algorithms that have been implemented in SpiNNaker.

Atomic force microscopy stiffness tomography on living Arabidopsis thaliana cells reveals the mechanical properties of surface and deep cell-wall layers during growth.

Science.gov (United States)

Radotić, Ksenija; Roduit, Charles; Simonović, Jasna; Hornitschek, Patricia; Fankhauser, Christian; Mutavdžić, Dragosav; Steinbach, Gabor; Dietler, Giovanni; Kasas, Sandor

2012-08-08

Cell-wall mechanical properties play a key role in the growth and the protection of plants. However, little is known about genuine wall mechanical properties and their growth-related dynamics at subcellular resolution and in living cells. Here, we used atomic force microscopy (AFM) stiffness tomography to explore stiffness distribution in the cell wall of suspension-cultured Arabidopsis thaliana as a model of primary, growing cell wall. For the first time that we know of, this new imaging technique was performed on living single cells of a higher plant, permitting monitoring of the stiffness distribution in cell-wall layers as a function of the depth and its evolution during the different growth phases. The mechanical measurements were correlated with changes in the composition of the cell wall, which were revealed by Fourier-transform infrared (FTIR) spectroscopy. In the beginning and end of cell growth, the average stiffness of the cell wall was low and the wall was mechanically homogenous, whereas in the exponential growth phase, the average wall stiffness increased, with increasing heterogeneity. In this phase, the difference between the superficial and deep wall stiffness was highest. FTIR spectra revealed a relative increase in the polysaccharide/lignin content. Copyright © 2012 Biophysical Society. Published by Elsevier Inc. All rights reserved.

Confined Electroconvective and Flexoelectric Instabilities Deep in the Freedericksz State of Nematic CB7CB.

Science.gov (United States)

Krishnamurthy, Kanakapura S; Palakurthy, Nani Babu; Yelamaggad, Channabasaveshwar V

2017-06-01

We report wormlike flexoelectric structures evolving deep in the Freedericksz state of a nematic layer of the liquid crystal cyanobiphenyl-(CH2) 7 -cyanobiphenyl. They form in the predominantly splay-bend thin boundary layers and are built up of solitary flexoelectric domains of the Bobylev-Pikin type. Their formation is possibly triggered by the gradient flexoelectric surface instability that remains optically discernible up to unusually high frequencies. The threshold voltage at which the worms form scales as square root of the frequency; in their extended state, worms often appear as labyrinthine structures on a section of loops that separate regions of opposite director deviation. Such asymmetric loops are also derived through pincement-like dissociation of ring-shaped walls. Formation of isolated domains of bulk electroconvection precedes the onset of surface instabilities. In essence, far above the Freedericksz threshold, the twisted nematic layer behaves as a combination of two orthogonally oriented planar half-layers destabilized by localized flexoelectric distortion.

Nitrogen Fertilizer Deep Placement for Increased Grain Yield and Nitrogen Recovery Efficiency in Rice Grown in Subtropical China

Directory of Open Access Journals (Sweden)

Meng Wu

2017-07-01

Full Text Available Field plot experiments were conducted over 3 years (from April 2014 to November 2016 in a double-rice (Oryza sativa L. cropping system in subtropical China to evaluate the effects of N fertilizer placement on grain yield and N recovery efficiency (NRE. Different N application methods included: no N application (CK; N broadcast application (NBP; N and NPK deep placement (NDP and NPKDP, respectively. Results showed that grain yield and apparent NRE significantly increased for NDP and NPKDP as compared to NBP. The main reason was that N deep placement (NDP increased the number of productive panicle per m-2. To further evaluate the increase, a pot experiment was conducted to understand the N supply in different soil layers in NDP during the whole rice growing stage and a 15N tracing technique was used in a field experiment to investigate the fate of urea-15N in the rice–soil system during rice growth and at maturity. The pot experiment indicated that NDP could maintain a higher N supply in deep soil layers than N broadcast for 52 days during rice growth. The 15N tracing study showed that NDP could maintain much higher fertilizer N in the 5–20 cm soil layer during rice growth and could induce plant to absorb more N from fertilizer and soil than NBP, which led to higher NRE. One important finding was that NDP and NPKDP significantly increased fertilizer NRE but did not lead to N declined in soil compared to NBP. Compared to NPK, NPKDP induced rice plants to absorb more fertilizer N rather than soil N.