Search NASA⌕ Search

SEARCH · Search NASA

Results for “deep data”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 757 records · Page 42

CROSS-CORRELATING COSMIC INFRARED AND X-RAY BACKGROUND FLUCTUATIONS: EVIDENCE OF SIGNIFICANT BLACK HOLE POPULATIONS AMONG THE CIB SOURCES

In order to understand the nature of the sources producing the recently uncovered cosmic infrared background (CIB) fluctuations, we study cross-correlations between the fluctuations in the source-subtracted CIB from Spitzer/IRAC data and the unresolved cosmic X-ray background from deep Chandra observations. Our study uses data from the EGS/AEGIS field, where both data sets cover an {approx_equal} 8' Multiplication-Sign 45' region of the sky. Our measurement is the cross-power spectrum between the IR and X-ray data. The cross-power signal between the IRAC maps at 3.6 {mu}m and 4.5 {mu}m and the Chandra [0.5-2] keV data has been detected, at angular scales {approx}> 20'', with an overall significance of {approx_equal} 3.8{sigma} and {approx_equal} 5.6{sigma}, respectively. At the same time we find no evidence of significant cross-correlations at the harder Chandra bands. The cross-correlation signal is produced by individual IR sources with 3.6 {mu}m and 4.5 {mu}m magnitudes m{sub AB} {approx}> 25-26 and [0.5-2] keV X-ray fluxes <<7 Multiplication-Sign 10{sup -17} erg cm{sup 2} s{sup -1}. We determine that at least 15%-25% of the large scale power of the CIB fluctuations is correlated with the spatial power spectrum of the X-ray fluctuations. If this correlation is attributed to emission from accretion processes at both IR and X-ray wavelengths, this implies a much higher fraction of accreting black holes than among the known populations. We discuss the various possible origins for the cross-power signal and show that neither local foregrounds nor the known remaining normal galaxies and active galactic nuclei can reproduce the measurements. These observational results are an important new constraint on theoretical modeling of the near-IR CIB fluctuations.

79 ASTRONOMY AND ASTROPHYSICS↗

SHARDS: A global view of the star formation activity at z ∼ 0.84 and z ∼ 1.23

In this paper, we present a comprehensive analysis of star-forming galaxies (SFGs) at intermediate redshifts (z ∼ 1). We combine the ultra-deep optical spectro-photometric data from the Survey for High-z Absorption Red and Dead Sources (SHARDS) with deep UV-to-FIR observations in the GOODS-N field. Exploiting two of the 25 SHARDS medium-band filters, F687W17 and F823W17, we select [O ii] emission line galaxies at z ∼ 0.84 and z ∼ 1.23 and characterize their physical properties. Their rest-frame equivalent widths (EW{sub rf}([O ii])), line fluxes, luminosities, star formation rates (SFRs), and dust attenuation properties are investigated. The evolution of EW{sub rf}([O ii]) closely follows the SFR density evolution of the universe, with a trend of EW{sub rf}([O ii]) ∝ (1 + z){sup 3} up to redshift z ≃ 1, followed by a possible flattening. The SF properties of the galaxies selected on the basis of their [O ii] emission are compared with complementary samples of SFGs selected by their MIR and FIR emission, and also with a general mass-selected sample of galaxies at the same redshifts. We demonstrate observationally that the UVJ diagram (or, similarly, a cut in the specific SFR) is only partially able to distinguish the quiescent galaxies from the SFGs. The SFR–M{sub *} relation is investigated for the different samples, yielding a logarithmic slope ∼1, in good agreement with previous results. The dust attenuations derived from different SFR indicators (UV(1600), UV(2800), [O ii], IR) are compared and show clear trends with respect to both the stellar mass and total SFR, with more massive and highly star-forming galaxies being affected by stronger dust attenuation.

79 ASTRONOMY AND ASTROPHYSICS↗

Randomized probe imaging through deep k-learning

Randomized probe imaging (RPI) is a single-frame diffractive imaging method that uses highly randomized light to reconstruct the spatial features of a scattering object. The reconstruction process, known as phase retrieval, aims to recover a unique solution for the object without measuring the far-field phase information. Typically, reconstruction is done via time-consuming iterative algorithms. In this work, we propose a fast and efficient deep learning based method to reconstruct phase objects from RPI data. The method, which we call deep k-learning, applies the physical propagation operator to generate an approximation of the object as an input to the neural network. This way, the network no longer needs to parametrize the far-field diffraction physics, dramatically improving the results. Deep k-learning is shown to be computationally efficient and robust to Poisson noise. The advantages provided by our method may enable the analysis of far larger datasets in photon starved conditions, with important applications to the study of dynamic phenomena in physical science and biological engineering.

Guo, Zhen (ORCID:0000000213473451)↗

Data Quality Monitoring for the Hadron Calorimeters Using Transfer Learning for Anomaly Detection

The proliferation of sensors brings an immense volume of spatio-temporal (ST) data in many domains, including monitoring, diagnostics, and prognostics applications. Data curation is a time-consuming process for a large volume of data, making it challenging and expensive to deploy data analytics platforms in new environments. Transfer learning (TL) mechanisms promise to mitigate data sparsity and model complexity by utilizing pre-trained models for a new task. Despite the triumph of TL in fields like computer vision and natural language processing, efforts on complex ST models for anomaly detection (AD) applications are limited. In this study, we present the potential of TL within the context of high-dimensional ST AD with a hybrid autoencoder architecture, incorporating convolutional, graph, and recurrent neural networks. Motivated by the need for improved model accuracy and robustness, particularly in scenarios with limited training data on systems with thousands of sensors, this research investigates the transferability of models trained on different sections of the Hadron Calorimeter of the Compact Muon Solenoid experiment at CERN. The key contributions of the study include exploring TL’s potential and limitations within the context of encoder and decoder networks, revealing insights into model initialization and training configurations that enhance performance while substantially reducing trainable parameters and mitigating data contamination effects.

47 OTHER INSTRUMENTATION↗

Graph neural networks for CO 2 solubility predictions in Deep Eutectic Solvents

Deep Eutectic Solvents (DESs) are a promising class of solvents for CO 2 capture. DESs are complex mixtures that can be designed to optimize CO solubility and overall capture process efficiency. However, the vast design landscape of DES mixtures makes experimental investigation prohibitive; as such, there is a need for computational models that can quickly and efficiently navigate the design space and inform data collection efforts. In this work, we propose Graph Neural Network (GNN) models for predicting CO 2 solubility for DESs; the GNN leverages a mixture graph representation that captures the molecular structure of the DES components as well as their intermolecular interactions. Here, we compare the GNN framework against alternative architectures (neural networks, graph convolution networks, and random forests) and data representations (molecular fingerprints, sigma profiles, and graphs). We show that the proposed approach offers superior predictive performance; specifically, we show that solubility can be predicted reliably directly from molecular structure (without the need of using sigma profiles as proposed in previous studies). This result is important, as obtaining sigma profiles requires expensive density functional theory computations. We also explored the ability of GNNs to predict solubility for new DES mixtures and operating conditions. We found that the model extrapolates across temperature reliably. However, we also found deficiencies in the ability of the model to predict solubility for DES mixtures, pressures, and molar ratio not included in the training sets; we show that this is due to an inherent lack of chemical diversity in datasets available in the literature. The proposed computational capabilities can thus help navigate the design space of DES and inform data collection efforts. Our models, data, and benchmarks are shared as Python code implemented in Jupyter notebooks.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Deep Learning for Unsupervised Insider Threat Detection in Structured Cyber Security Data Streams

Analysis of an organization's computer network activity is a key component of early detection and mitigation of insider threat, a growing concern for many organizations. Raw system logs are a prototypical example of streaming data that can quickly scale beyond the cognitive power of a human analyst. As a prospective filter for the human analyst, we present an online unsupervised deep learning approach to detect anomalous network activity from system logs in real time. Our models decompose anomaly scores into the contributions of individual user behavior features for increased interpretability to aid analysts reviewing potential cases of insider threat. Using the CERT Insider Threat Dataset v6.2 and threat detection recall as our performance metric, our novel deep and recurrent neural network models outperform Principal Component Analysis, Support Vector Machine and Isolation Forest based anomaly detection baselines. For our best model, the events labeled as insider threat activity in our dataset had an average anomaly score in the 95.53 percentile, demonstrating our approach's potential to greatly reduce analyst workloads.

Deep Learning, Insider Threat↗

Deep multiscale model learning

The objective of this article is to design novel multi-layer neural networks for multiscale simulations of flows taking into account the observed fine data and physical modeling concepts. Our approaches use deep learning techniques combined with local multiscale model reduction methodologies to predict flow dynamics. Using reduced-order model concepts is important for constructing robust deep learning architectures since the reduced-order models provide fewer degrees of freedom. We consider flow dynamics in porous media as multi-layer networks in this work. More precisely, the solution (e.g., pressures and saturation) at the time instant n+1 depends on the solution at the time instant n and input parameters, such as permeability fields, forcing terms, and initial conditions. One can regard the solution as a multi-layer network, where each layer, in general, is a nonlinear forward map and the number of layers relates to the internal time steps. We will rely on rigorous model reduction concepts to define unknowns and connections between layers. It is critical to use reduced-order models for this purpose, which will identify the regions of influence and the appropriate number of variables. Furthermore, due to the lack of available observed fine data, the reduced-order model can provide us sufficient inexpensive data as needed. The designed deep neural network will be trained using both coarse simulation data which is obtained from the reduced-order model and observed fine data. We will present the main ingredients of our approach and numerical examples. Numerical results show that using deep learning with data generated from multiscale models as well as available observed fine data, we can obtain an improved forward map which can better approximate the fine scale model.

97 MATHEMATICS AND COMPUTING↗

( Chandra Chandra) counterparts of CANDELS GOODS-S sources

Improving the capabilities of detecting faint X-ray sources is fundamental for increasing the statistics on faint high-z active galactic nuclei (AGNs) and star-forming galaxies (SFGs). We performed a simultaneous maximum likelihood point-spread function fit in the [0.5–2] keV and [2–7] keV energy bands of the 4 Ms Chandra Deep Field South (CDFS) data at the position of the 34,930 CANDELS H-band selected galaxies. For each detected source we provide X-ray photometry and optical counterpart validation. We validated this technique by means of a ray-tracing simulation. We detected a total of 698 X-ray point sources with a likelihood L>4.98 (i.e., >2.7σ). We show that prior knowledge of a deep sample of optical–NIR galaxies leads to a significant increase in the detection of faint (i.e., ∼10{sup −17} cgs in the [0.5–2] keV band) sources with respect to “blind” X-ray detections. By including previous X-ray catalogs, this work increases the total number of X-ray sources detected in the 4 Ms CDFS, CANDELS area to 793, which represents the largest sample of extremely faint X-ray sources assembled to date. Our results suggest that a large fraction of the optical counterparts of our X-ray sources determined by likelihood ratio actually coincides with the priors used for the source detection. Most of the new detected sources are likely SFGs or faint, absorbed AGNs. We identified a few sources with putative photometric redshift z > 4. Despite the low number statistics and the uncertainties on the photo z, this sample significantly increases the number of X-ray-selected candidate high-z AGNs.

79 ASTRONOMY AND ASTROPHYSICS↗

AI for Science: Report on the Department of Energy (DOE) Town Halls on Artificial Intelligence (AI) for Science

The report documents the DOE Town Halls held during 2019 at Argonne National Laboratory, Oak Ridge National Laboratory, Lawrence Berkeley National Laboratory, and in Washington, DC. From July to October 2019, the Argonne, Oak Ridge, and Berkeley National Laboratories hosted a series of four town hall meetings attended by more than 1,000 U.S. scientists and engineers. The goal of the town hall series was to examine scientific opportunities in the areas of artificial intelligence (AI), Big Data, and high-performance computing (HPC) in the next decade, and to capture the big ideas, grand challenges, and next steps to realizing these opportunities. In this report and in the Department of Energy (DOE) laboratory community, we use the term “AI for Science” to broadly represent the next generation of methods and scientific opportunities in computing, including the development and application of AI methods (e.g., machine learning, deep learning, statistical methods, data analytics, automated control, and related areas) to build models from data and to use these models alone or in conjunction with simulation and scalable computing to advance scientific research. The AI for Science town hall discussions focused on capturing the transformational uses of AI that employ HPC and/or data analysis, leveraging data sets from HPC simulations or instruments and user facilities, and addressing scientific challenges unique to DOE user facilities and the agency’s wide-ranging fundamental and applied science enterprise.

36 MATERIALS SCIENCE↗

Semi-inclusive DIS cross sections and spin asymmetries in the quantum statistical parton distributions approach

We consider the extension of the statistical parton distributions to include their transverse momentum dependence, by using two different methods, one is based on our quantum statistical approach, the other on a relativistic covariant method. We take into account the effects of the Melosh-Wigner rotation for the polarized distributions. The results obtained can be compared with recent semi-inclusive deep inelastic scattering (DIS) data on the cross section and double longitudinal-spin asymmetries from JLab. We also give some predictions for future experiments on electron-neutron scattering.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Smart Scattering Scanning Near-Field Optical Microscopy

Scattering scanning near-field optical microscopy (s-SNOM) provides spectroscopic imaging from molecular to quantum materials with few nanometer deep subdiffraction limited spatial resolution. However, in its conventional implementation s-SNOM is slow to effectively acquire a series of spatio-spectral images, especially with large fields of view. This problem is further exacerbated for weak resonance contrast or when using light sources with limited spectral irradiance. Indeed, the generally limited signal-to-noise ratio prevents sampling a weak signal at the Nyquist sampling rate. Here, we demonstrate how acquisition time and sampling rate can be significantly reduced by using compressed sampling, matrix completion, and adaptive random sampling, while maintaining or even enhancing the physical or chemical image content. We use fully sampled real data sets of molecular, biological, and quantum materials as ground-truth physical data and show how deep under-sampling with a corresponding reduction of acquisition time by 1 order of magnitude or more retains the core s-SNOM image information. We demonstrate that a sampling rate of up to 6× smaller than the Nyquist criterion can be applied, which would provide a 30-fold reduction in the data required under typical experimental conditions. Furthermore, our smart s-SNOM approach is generally applicable and provides systematic full spatio-spectral s-SNOM imaging with a large field of view at high spectral resolution and reduced acquisition time.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Providing Multi-Page Data Extraction Services with XWRAPComposer

Dynamic Web data sources – sometimes known collectively as the Deep Web – increase the utility of the Web by providing intuitive access to data repositories anywhere that Web access is available. Deep Web services provide access to real-time information, like entertainment event listings, or present a Web interface to large databases or other data repositories. Recent studies suggest that the size and growth rate of the dynamic Web greatly exceed that of the static Web, yet dynamic content is often ignored by existing search engine indexers owing to the technical challenges that arise when attempting to search the Deep Web. To address these challenges, we present DYNABOT, a service-centric crawler for discovering and clustering Deep Web sources offering dynamic content. DYNABOT has three unique characteristics. First, DYNABOT utilizes a service class model of the Web implemented through the construction of service class descriptions (SCDs). Second, DYNABOT employs a modular, self-tuning system architecture for focused crawling of the Deep Web using service class descriptions. Third, DYNABOT incorporates methods and algorithms for efficient probing of the Deep Web and for discovering and clustering Deep Web sources and services through SCD-based service matching analysis. Our experimental results demonstrate the effectiveness of the service class discovery, probing, and matching algorithms and suggest techniques for efficiently managing service discovery in the face of the immense scale of the Deep Web.

97 MATHEMATICS AND COMPUTING↗

Deep Learning for Full Waveform Inversion of Elastic Active-Source Seismic Data to Estimate P-Wave Velocity Models

Seismic imaging methods are critical for Global Security and Energy & Homeland Security missions and activities that rely on subsurface characterization, but traditional methods remain computationally expensive and require significant labor hours and expertise to execute. Within the past few years, machine learning (ML), namely deep learning (DL), has been used to develop data-driven end-to-end full waveform inversion (FWI) methods to estimate 2D P-wave velocity (Vp) models in a fraction of the time as conventional FWI. These methods, however, are trained on simplistic acoustic wave seismic data and Vp models that are not realistic nor representative of real-world observations, leaving a large gap between the state-of-the-art and deployable, feasible, and practical DL FWI methods. Here, we generate a synthetic active-source, 3D, elastic wave seismic data set and a variety of Vp models with realistic geologic structure for training DL FWI methods. We evaluate six different methods that have performed well for acoustic DL FWI or medical imaging tasks using our more realistic dataset. We find that these six trained models do not match the performance of published acoustic end-to-end DL FWI methods, indicating more training data may be needed, physics may need to be incorporated to achieve good accuracy at the sacrifice of the end-to-end advantage, and/or novel methods need to be developed to enable end-to-end DL FWI methods to perform well for real-world seismic data.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

THE BRIGHT END OF THE ULTRAVIOLET LUMINOSITY FUNCTION AT z {approx} 8: NEW CONSTRAINTS FROM CANDELS DATA IN GOODS-SOUTH

We present new z {approx} 8 galaxy candidates from a search over {approx}95 arcmin{sup 2} of WFC3/IR data, tripling the previous search area for bright z {approx} 8 galaxies. Our analysis uses newly acquired WFC3/IR imaging data from the CANDELS Multi-Cycle Treasury program over the GOODS-South field. These new data are combined with existing deep optical Advanced Camera for Surveys (ACS) imaging to search for relatively bright (M {sub UV} < -19.5 mag) z {approx} 8 galaxy candidates using the Lyman break technique. These new candidates are used to determine the bright end of the UV luminosity function (LF) of star-forming galaxies at z {approx} 7.2-8.7, i.e., a cosmic age of 600 {+-} 80 Myr. To minimize contamination from lower redshift galaxies, we make full use of all optical ACS data and impose strict non-detection criteria based on an optical {chi}{sup 2} {sub opt} flux measurement. In the whole search area, we identify 16 candidate z {approx} 8 galaxies, spanning a magnitude range H {sub 160,AB} = 25.7-27.9 mag. The new data show that the UV LF is a factor {approx}1.7 lower at M {sub UV} < -19.5 mag than determined from the HUDF09 and Early Release Science (ERS) data alone. Combining this new sample with the previous candidates from the HUDF09 and ERS data allows us to perform the most accurate measurement of the z {approx} 8 UV LF yet. Schechter function fits to the combined data result in a best-fit characteristic magnitude of M {sub *}(z = 8) = -20.04 {+-} 0.46 mag. The faint-end slope is very steep, though quite uncertain, with {alpha} = -2.06 {+-} 0.32. A combination of wide-area data with additional ultra-deep imaging will be required to significantly reduce the uncertainties on these parameters in the future.

79 ASTRONOMY AND ASTROPHYSICS↗

The X-ray luminosity functions of field low-mass X-ray binaries in early-type galaxies: Evidence for a stellar age dependence

We present direct constraints on how the formation of low-mass X-ray binary (LMXB) populations in galactic fields depends on stellar age. In this pilot study, we utilize Chandra and Hubble Space Telescope (HST) data to detect and characterize the X-ray point source populations of three nearby early-type galaxies: NGC 3115, 3379, and 3384. The luminosity-weighted stellar ages of our sample span ≈3-10 Gyr. X-ray binary population synthesis models predict that the field LMXBs associated with younger stellar populations should be more numerous and luminous per unit stellar mass than older populations due to the evolution of LMXB donor star masses. Crucially, the combination of deep Chandra and HST observations allows us to test directly this prediction by identifying and removing counterparts to X-ray point sources that are unrelated to the field LMXB populations, including LMXBs that are formed dynamically in globular clusters, Galactic stars, and background active galactic nuclei/galaxies. We find that the 'young' early-type galaxy NGC 3384 (≈2-5 Gyr) has an excess of luminous field LMXBs (L {sub X} ≳ (5-10) × 10{sup 37} erg s{sup –1}) per unit K-band luminosity (L{sub K} ; a proxy for stellar mass) than the 'old' early-type galaxies NGC 3115 and 3379 (≈8-10 Gyr), which results in a factor of ≈2-3 excess of L {sub X}/L{sub K} for NGC 3384. This result is consistent with the X-ray binary population synthesis model predictions; however, our small galaxy sample size does not allow us to draw definitive conclusions on the evolution field LMXBs in general. We discuss how future surveys of larger galaxy samples that combine deep Chandra and HST data could provide a powerful new benchmark for calibrating X-ray binary population synthesis models.

79 ASTRONOMY AND ASTROPHYSICS↗

Intercomparison of Deep Learning Model Architectures for Atmospheric River Prediction

With a rapid surge in the application of machine learning (ML) for a diverse range of tasks in climate science, the present study addresses a challenge for climate scientists when selecting the optimal ML or deep learning (DL) architecture for a given application. In particular, a DL intercomparison study was performed with a focus on forecasting the position of atmospheric rivers (ARs) on short-range time scales (up to 5-day lead times). AR predictions from multiple DL architectures, including various types of convolutional autoencoders and a vision transformer (ViT), were compared against ECMWF ERA5 reanalysis and hindcasts from a global climate model. DL models with similar trainable parameters were trained on ERA5 reanalysis data and AR positions derived from a thresholding algorithm to ensure a fair comparison among the DL models. Each model’s performance and accuracy in forecasting AR location and key input fields within a 5-day window were assessed using metrics of root-mean-square error, anomaly correlation, and mean intersection over union. The ViT architecture outperformed other autoencoder models in most of the metrics. Incorporating additional meteorological fields only yielded slight improvements in forecasting certain fields at longer lead times. The results also suggest that a smaller number of input time steps or smaller number of autoregressive steps can achieve better prediction skills, while also improving the overall computational efficiency. This research offers valuable insights into the strengths and weaknesses of different DL techniques for AR forecasting, hopefully guiding the development of improved models for forecasting this phenomenon.

54 ENVIRONMENTAL SCIENCES↗

Using Ultrasound Image Augmentation and Ensemble Predictions to Prevent Machine-Learning Model Overfitting

Deep learning predictive models have the potential to simplify and automate medical imaging diagnostics by lowering the skill threshold for image interpretation. However, this requires predictive models that are generalized to handle subject variability as seen clinically. Here, we highlight methods to improve test accuracy of an image classifier model for shrapnel identification using tissue phantom image sets. Using a previously developed image classifier neural network—termed ShrapML—blind test accuracy was less than 70% and was variable depending on the training/test data setup, as determined by a leave one subject out (LOSO) holdout methodology. Introduction of affine transformations for image augmentation or MixUp methodologies to generate additional training sets improved model performance and overall accuracy improved to 75%. Further improvements were made by aggregating predictions across five LOSO holdouts. This was done by bagging confidences or predictions from all LOSOs or the top-3 LOSO confidence models for each image prediction. Top-3 LOSO confidence bagging performed best, with test accuracy improved to greater than 85% accuracy for two different blind tissue phantoms. This was confirmed by gradient-weighted class activation mapping to highlight that the image classifier was tracking shrapnel in the image sets. Overall, data augmentation and ensemble prediction approaches were suitable for creating more generalized predictive models for ultrasound image analysis, a critical step for real-time diagnostic deployment.

60 APPLIED LIFE SCIENCES↗