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 469 records · Page 26

RC-SFA Data Management Templates and Guidance for Standardized, Reusable AI-Ready Data Packages

This data package provides templates and supporting documentation developed by the River Corridor Science Focus Area (RC-SFA; https://www.pnnl.gov/projects/river-corridor) to communicate its approach to managing and publishing AI-ready data. The package is intended to help data users and data producers understand the structures, metadata practices, and quality-control approaches that support consistent, reusable, and machine-actionable data products across RC-SFA studies. Rather than focusing on a single experimental dataset, this package documents the data management framework used to make RC-SFA data easier to find, ingest, navigate, and interpret. The materials in this package reflect RC-SFA practices for standardized data package organization, including the use of a human- and machine-readable README, file-level metadata, data dictionaries, descriptive file naming, method identifiers, and automated and review-based quality assurance procedures. Together, these components illustrate how RC-SFA extends FAIR data principles toward AI-readiness by prioritizing deep metadata, consistency across data packages, and support for informed downstream reuse by both humans and computational tools. This dataset is comprised of (1) readme; (2) presentation slides with an overview of RC-SFA approach and guidance; (3) document of RC-SFA best practices; (4) data dictionary (dd); (5) file level metadata (flmd); and a subfolder containing templates for dd and flmd. All files are .csv and .pdf. For details on how to navigate data packages generated by this project, see https://data.ess-dive.lbl.gov/portals/PNNLRiverCorridorSFA/About.

AI-readiness↗

How Low Can You Go? Using Synthetic 3D Imagery to Drastically Reduce Real-World Training Data for Object Detection

Deep convolutional neural networks (DCNNs) currently provide state-of-the-art performance on image classification and object detection tasks, and there are many global security mission areas where such models could be extremely useful. Crucially, the success of these models is driven in large part by the widespread availability of high-quality open source data sets such as Image Net, Common Objects in Context (COCO), and KITTI, which contain millions of images with thousands of unique labels. However, global security relevant objects-of-interest can be difficult to obtain: relevant events are low frequency and high consequence; the content of relevant images is sensitive; and adversaries and proliferators seek to obscure their activities. For these cases where exemplar data is hard to come-by, even fine-tuning an existing model with available data can be effectively impossible. Recent work demonstrated that models can be trained using a combination of real-world and synthetic images generated from 3D representations; that such models can exceed the performance of models trained using real-world data alone; and that the generated images need not be perfectly realistic (Tremblay, et al., 2018). However, this approach still required hundreds to thousands of real-world images for training and fine tuning, which for sparse, global security-relevant datasets can be an unrealistic hurdle. In this research, we validate the performance and behavior of DCNN models as we drive the number of real-world images used for training object detection tasks down to a minimal set. We perform multiple experiments to identify the best approach to train DCNNs from an extremely small set of real-world images. In doing so, we: Develop state-of-the-art, parameterized 3D models based on real-world images and sample from their parameters to increase the variance in synthetic image training data; Use machine learning explainability techniques to highlight and correct through targeted training the biases that result from training using completely synthetic images; and Validate our results by comparing the performance of the models trained on synthetic data to one another, and to a control model created by fine-tuning an existing ImageNet-trained model with a limited number (hundreds) of real-world images.

97 MATHEMATICS AND COMPUTING↗

A VERY COMPACT DENSE GALAXY OVERDENSITY WITH δ ≃ 130 IDENTIFIED AT z ∼ 8: IMPLICATIONS FOR EARLY PROTOCLUSTER AND CLUSTER CORE FORMATION

We report the first identification of a compact dense galaxy overdensity at z∼8 called A2744z8OD. A2744z8OD consists of eight Y-dropout galaxies behind Abell 2744 that were originally pinpointed by Hubble Frontier Fields studies. However, so far, no studies have derived the basic physical quantities of structure formation or made comparisons with theoretical models. We obtain a homogeneous sample of dropout galaxies at z∼8 from eight field data of Hubble legacy images that are as deep as the A2744z8OD data. Using the sample, we find that a galaxy surface overdensity value of A2744z8OD is very high δ≃130, where δ is defined by an overdensity in a small circle of 6″ (≃30 physical kiloparsecs) radius. Because there is no such large δ value reported for high-z overdensities to date, A2744z8OD is a system that is clearly different from those found in previous high-z overdensity studies. In the galaxy+structure formation models of Henriques et al., there exists a very similar overdensity, Modelz8OD, that is made of eight model dropout galaxies at z∼8 in a 6″ radius circle. Modelz8OD is a progenitor of a today’s 10{sup 14}M{sub ⊙} cluster, and more than half of the seven Modelz8OD galaxies are merged into the brightest cluster galaxy of the cluster. If Modelz8OD is a counterpart to A2744z8OD, the models suggest that A2744z8OD would be part of a cluster core forming from a 10{sup 14}M{sub ⊙} cluster that began star formation at z>12.

79 ASTRONOMY AND ASTROPHYSICS↗

Geometric scaling at BNL RHIC and CERN LHC

We present a new phenomenological model of the dipole scattering amplitude to demonstrate that the RHIC data for hadron production in d-Au collisions for all available rapidities are compatible with geometric scaling, just like the small-x inclusive deep inelastic scattering data. A detailed comparison with earlier geometric scaling violating models of the dipole scattering amplitude in terms of an anomalous dimension {gamma} is made. In order to establish whether the geometric scaling violations expected from small-x evolution equations are present in the data a much larger range in transverse momentum and rapidity must be probed. Predictions for hadron production in p-Pb and p-p collisions at LHC are given. We point out that the falloff of the transverse momentum distribution at LHC is a sensitive probe of the variation of {gamma} in a region where x is much smaller than at RHIC. In this way, the expectation for the rise of {gamma} from small-x evolution can be tested.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Attention-based functional-group coarse-graining: a deep learning framework for molecular prediction and design

Machine learning (ML) offers considerable promise for the design of new molecules and materials. In real-world applications, the design problem is often domain-specific, and suffers from insufficient data, particularly labeled data, for ML training. In this study, we report a data-efficient, deep-learning framework for molecular discovery that integrates a coarse-grained functional-group representation with a self-attention mechanism to capture intricate chemical interactions. Our approach exploits group-contribution concepts to create a graph-based intermediate representation of molecules, serving as a low-dimensional embedding that substantially reduces the data demands typically required for training. Using a self-attention mechanism to learn the subtle but highly relevant chemical context of functional groups, the method proposed here consistently outperforms existing approaches for predictions of multiple thermophysical properties. In a case study focused on adhesive polymer monomers, we train on a limited dataset comprising only 6,000 unlabeled and 600 labeled monomers. The resulting chemistry prediction model achieves over 92% accuracy in forecasting properties directly from SMILES strings, exceeding the performance of current state-of-the-art techniques. Furthermore, the latent molecular embedding is invertible, enabling the design pipeline to automatically generate new monomers from the learned chemical subspace. We illustrate this functionality by targeting several properties, including high and low glass transition temperatures (Tg), and demonstrate that our model can identify new candidates with values that surpass those in the training set. The ease with which the proposed framework navigates both chemical diversity and data scarcity offers a promising route to accelerate and broaden the search for functional materials.

Han, Ming [Univ. of Chicago, IL (United States)]↗

Deep learning ferroelectric polarization distributions from STEM data via with and without atom finding

Over the last decade, scanning transmission electron microscopy (STEM) has emerged as a powerful tool for probing atomic structures of complex materials with picometer precision, opening the pathway toward exploring ferroelectric, ferroelastic, and chemical phenomena on the atomic scale. Analyses to date extracting a polarization signal from lattice coupled distortions in STEM imaging rely on discovery of atomic positions from intensity maxima/minima and subsequent calculation of polarization and other order parameter fields from the atomic displacements. Here, we explore the feasibility of polarization mapping directly from the analysis of STEM images using deep convolutional neural networks (DCNNs). In this approach, the DCNN is trained on the labeled part of the image (i.e., for human labelling), and the trained network is subsequently applied to other images. We explore the effects of the choice of the descriptors (centered on atomic columns and grid-based), the effects of observational bias, and whether the network trained on one composition can be applied to a different one. This analysis demonstrates the tremendous potential of the DCNN for the analysis of high-resolution STEM imaging and spectral data and highlights the associated limitations.

36 MATERIALS SCIENCE↗

New parton distributions for collider physics

We extract new parton distribution functions (PDFs) of the proton by global analysis of hard scattering data in the general-mass framework of perturbative quantum chromodynamics. Our analysis includes new theoretical developments together with the most recent collider data from deep-inelastic scattering, vector boson production, and single-inclusive jet production. Because of the difficulty in fitting both the D0 Run-II W lepton asymmetry data and some fixed-target DIS data, we present two families of PDFs, CT10 and CT10W, without and with these high-luminosity W lepton asymmetry data included in the global analysis. With both sets of PDFs, we study theoretical predictions and uncertainties for a diverse selection of processes at the Fermilab Tevatron and the CERN Large Hadron Collider.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Molecular Simulation of Lithium Carbonate Reactive Vapor–Liquid Equilibria Using a Deep Potential Model

We developed a first-principles machine learning model for the reactive vapor–liquid phase behavior of molten Li 2 CO 3 . The model was trained on ab initio electronic density functional theory data using the Deep Potential (DP) methodology, and its accuracy was evaluated by comparing model predictions of density and viscosity to experimental measurements. Direct coexistence simulations with the DP model over time scales of tens of nanoseconds were used to observe equilibrium dissociation of Li 2 CO 3 into CO 2 residing primarily in the vapor phase and Li 2 O which remains dissolved in the liquid. The simulations covered a range of temperatures, overall system sizes, and vapor-to-liquid volume ratios. Results were analyzed in terms of the observed chemical composition of the liquid and vapor phases, product structure, and CO 2 partial pressures. In addition, we calculated equilibrium constants for the dissociation reaction by assuming ideal-solution behavior for the liquid. As expected on the basis of thermodynamic arguments and prior experiments for this system, the observed partial pressure of CO 2 in the gas phase depends on both the temperature and the ratio of vapor to liquid volumes, while the calculated equilibrium constants only depend on temperature. DP model predictions for the equilibrium constant of the reaction are generally consistent with the available experimental measurements. Furthermore, the present study establishes the validity of the DP methodology for the description of reactive, multiphase equilibria from first principles, with possible applications to many other systems of scientific and technological interest even in the absence of relevant experimental measurements.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A physics-informed operator regression framework for extracting data-driven continuum models

The application of deep learning toward discovery of data-driven models requires careful application of inductive biases to obtain a description of physics which is both accurate and robust. We present here a framework for discovering continuum models from high fidelity molecular simulation data. Our approach applies a neural network parameterization of governing physics in modal space, allowing a characterization of differential operators while providing structure which may be used to impose biases related to symmetry, isotropy, and conservation form. Here, we demonstrate the effectiveness of our framework for a variety of physics, including local and nonlocal diffusion processes and single and multiphase flows. For the flow physics we demonstrate this approach leads to a learned operator that generalizes to system characteristics not included in the training sets, such as variable particle sizes, densities, and concentration.

42 ENGINEERING↗

Using next generation transcriptome sequencing to predict an ectomycorrhizal metablome.

Mycorrhizae, symbiotic interactions between soil fungi and tree roots, are ubiquitous in terrestrial ecosystems. The fungi contribute phosphorous, nitrogen and mobilized nutrients from organic matter in the soil and in return the fungus receives photosynthetically-derived carbohydrates. This union of plant and fungal metabolisms is the mycorrhizal metabolome. Understanding this symbiotic relationship at a molecular level provides important contributions to the understanding of forest ecosystems and global carbon cycling. We generated next generation short-read transcriptomic sequencing data from fully-formed ectomycorrhizae between Laccaria bicolor and aspen (Populus tremuloides) roots. The transcriptomic data was used to identify statistically significantly expressed gene models using a bootstrap-style approach, and these expressed genes were mapped to specific metabolic pathways. Integration of expressed genes that code for metabolic enzymes and the set of expressed membrane transporters generates a predictive model of the ectomycorrhizal metabolome. The generated model of mycorrhizal metabolome predicts that the specific compounds glycine, glutamate, and allantoin are synthesized by L. bicolor and that these compounds or their metabolites may be used for the benefit of aspen in exchange for the photosynthetically-derived sugars fructose and glucose. The analysis illustrates an approach to generate testable biological hypotheses to investigate the complex molecular interactions that drive ectomycorrhizal symbiosis. These models are consistent with experimental environmental data and provide insight into the molecular exchange processes for organisms in this complex ecosystem. The method used here for predicting metabolomic models of mycorrhizal systems from deep RNA sequencing data can be generalized and is broadly applicable to transcriptomic data derived from complex systems.

54 ENVIRONMENTAL SCIENCES↗

Granite Springs Valley, Nevada Play Fairway Analysis - Well data and Temperature Survey

This data is associated with the Nevada Play Fairway project and includes excel files containing raw 2-meter temperature data and corrections. GIS shapefiles and layer files contain ing location and attribute information for the data are included. Well data includes both deep and shallow TG holes, GIS shapefiles and layer files.

15 GEOTHERMAL ENERGY↗

Meson Form Factors and Deep Exclusive Meson Production Experiments

Pion and kaon electroproduction data play a unique role in Nature and our understanding of them is essential for explaining hadron structure. Precision longitudinaltransverse separated pion and kaon cross sections are of particular interest. They allow for the extraction of meson form factors and validation of understanding of hard exclusive and semi-inclusive reactions (π+, K+, π0, γ) towards 3D hadron imaging and potential future flavor decomposition. We review recent data and present prospects for deep exclusive pion and kaon electroproduction at the 12 GeV Jefferson Lab including the prospects to use projected charged- and neutral pion data to further determine the spin, charge-parity and flavor of GPDs, including the helicity-flip GPDs.

Horn, Tanja↗

Deep-Learning-derived Boundary Layer Height from Meteorological Data over the SGP, GOAMAZON, CACTI

The planetary boundary-layer (PBL) height (PBLH) is an important parameter for various meteorological and climate studies. This study presents a multi-structure deep neural network (DNN) model, designed to estimate PBLH by integrating morning temperature profiles with surface meteorological observations. The DNN model is developed by leveraging a rich data set of PBLH derived from long-standing radiosonde records and augmented with high-resolution micropulse lidar and Doppler lidar observations. We access the performance of the DNN with an ensemble of 10 members, each featuring distinct hidden layer structures, which collectively yield a robust 27-year PBLH data set over the Southern Great Plains from 1994 to 2020. The influence of various meteorological factors on PBLH is rigorously analyzed through the importance test. Moreover, the DNN model's accuracy is evaluated against radiosonde observations and juxtaposed with conventional remote-sensing methodologies, including Doppler lidar, ceilometer, Raman lidar, and micropulse lidar. The DNN model exhibits reliable performance across diverse conditions and demonstrates lower biases relative to remote-sensing methods. In addition, the DNN model, originally trained over a plain region, demonstrates remarkable adaptability when applied to the heterogeneous terrains and climates encountered during the GoAmazon (tropical rainforest) and CACTI (middle-latitude mountain) campaigns. These findings demonstrate the effectiveness of deep learning models in estimating PBLH, enhancing our understanding of boundary-layer dynamics with implications for enhancing the representation of PBL in weather forecasting and climate modeling.

54 ENVIRONMENTAL SCIENCES↗

Deep Generative Models that Solve PDEs: Distributed Computing for Training Large Data-Free Models

Recent progress in scientific machine learning (SciML) has opened up the possibility of training novel neural network architectures that solve complex partial differential equations (PDEs). Several (nearly data free) approaches have been recently reported that successfully solve PDEs, with examples including deep feed forward networks, generative networks, and deep encoder-decoder networks. However, practical adoption of these approaches is limited by the difficulty in training these models, especially to make predictions at large output resolutions (≥1024×1024). Here we report on a software framework for data parallel distributed deep learning that resolves the twin challenges of training these large SciML models - training in reasonable time as well as distributing the storage requirements. Our framework provides several out of the box functionality including (a) loss integrity independent of number of processes, (b) synchronized batch normalization, and (c) distributed higher-order optimization methods. We show excellent scalability of this framework on both cloud as well as HPC clusters, and report on the interplay between bandwidth, network topology and bare metal vs cloud. We deploy this approach to train generative models of sizes hitherto not possible, showing that neural PDE solvers can be viably trained for practical applications. We also demonstrate that distributed higher-order optimization methods are 2-3× faster than stochastic gradient-based methods and provide minimal convergence drift with higher batch-size.

PDEs↗

Impact of W and Z Production Data and Compatibility of Neutrino DIS Data in Nuclear Parton Distribution Functions

Vector boson production and neutrino deep-inelastic scattering (DIS) data are crucial for constraining the strange quark parton distribution function (PDF) and more generally for flavor decomposition in PDF extractions. We extend the nCTEQ15 nuclear PDFs (nPDFs) by adding the recent W and Z production data from the LHC in a global nPDF fit. The new nPDF set, referred to as nCTEQ15WZ, is used as a starting point for a follow-up study in which we assess the compatibility of neutrino DIS data with charged lepton DIS data. Specifically, we re-analyze neutrino DIS data from NuTeV, Chorus, and CDHSW, as well as dimuon data from CCFR and NuTeV. To scrutinize the level of compatibility, different kinematic regions of the neutrino data are investigated. Fits to the neutrino data alone and a preliminary global fit are performed and compared to nCTEQ15WZ.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Design of the protoDUNE raw data management infrastructure

The Deep Underground Neutrino Experiment (DUNE) will employ a set of Liquid Argon Time Projection Chambers (LArTPC) with a total mass of 40 kt as the main components of its Far Detector. In order to validate this technology and characterize the detector performance at full scale, an ambitious experimental program (called “protoDUNE”) has been initiated which includes a test of the large-scale prototypes for the single-phase and dual-phase LArTPC technologies, which will run in a beam at CERN. The total raw data volume that is slated to be collected during the scheduled 3-month beam run is estimated to be in excess of 2.5 PB for each detector. This data volume will require that the protoDUNE experiment carefully design the DAQ, data handling and data quality monitoring systems to be capable of dealing with challenges inherent with peta-scale data management while simultaneously fulfilling the requirements of disseminating the data to a worldwide collaboration and DUNE associated computing sites. Here in this paper, we present our approach to solving these problems by leveraging the design, expertise and components created for the LHC and Intensity Frontier experiments into a unified architecture that is capable of meeting the needs of protoDUNE.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Hawthorne Nevada Deep Direct-Use Feasibility Study - Geophysics GIS Data

The objective of this project is to use a multi-disciplinary, three-tiered approach to assess the geothermal resource and determine the feasibility of implementing a large-scale, direct-use facility for the Hawthorne Army Depot (HAD) and the various town and county facilities in Hawthorne, Nevada. This assessment directly targets a geothermal resource recently characterized by the Navy Geothermal Program Office (GPO) as part of a focused exploration and development campaign. This dataset includes the GIS data related to geophysics.

15 GEOTHERMAL ENERGY↗