SEARCH · Search NASA

Results for “DL”

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.

SourceResearch connectionFrom yearThrough year

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

At least 91 records · Page 5

All sources NASA NTRS DOE OSTI

DOE OSTI2022

Anderson Acceleration for Distributed Training of Deep Learning Models

Anderson acceleration (AA) is an extrapolation technique that has recently gained interest in the deep learning (DL) community to speed-up the sequential training of DL models. However, when performed at large scale, the DL training is exposed to a higher risk of getting trapped into steep local minima of the training loss function, and standard AA does not provide sufficient acceleration to escape from these steep local minima. This results in poor generalizability and makes AA ineffective. To restore AA’s advantage to speed-up the training of DL models on large scale computing platforms, we combine AA with an adaptive moving average procedure that boosts the training to escape from steep local minima. By monitoring the relative standard deviation between consecutive iterations, we also introduce a criterion to automatically assess whether the moving average is needed. We applied the method to the following DL instantiations for image classification: (i) ResNet50 trained on the open-source CIFAR100 dataset and (ii) ResNet50 trained on the open-source ImageNet1k dataset. Numerical results obtained using up to 1,536 NVIDIA V100 GPUs on the OLCF supercomputer Summit showed the stabilizing effect of the moving average on AA for all the problems above.

Lupo Pasini, Massimiliano↗

NASA NTRS2023

Explainable deep learning for insights in El Niño and river flows

The El Niño Southern Oscillation (ENSO) is a semi-periodic fluctuation in sea surface temperature (SST) over the tropical central and eastern Pacific Ocean that influences interannual variability in regional hydrology across the world through long-range dependence or teleconnections. Recent research has demonstrated the value of Deep Learning (DL) methods for improving ENSO prediction as well as Complex Networks (CN) for understanding teleconnections. However, gaps in predictive understanding of ENSO-driven river flows include the black box nature of DL, the use of simple ENSO indices to describe a complex phenomenon and translating DL-based ENSO predictions to river flow predictions. Here we show that eXplainable DL (XDL) methods, based on saliency maps, can extract interpretable predictive information contained in global SST and discover SST information regions and dependence structures relevant for river flows which, in tandem with climate network constructions, enable improved predictive understanding. Our results reveal additional information content in global SST beyond ENSO indices, develop understanding of how SSTs influence river flows, and generate improved river flow prediction, including uncertainty estimation. Observations, reanalysis data, and earth system model simulations are used to demonstrate the value of the XDL-CN based methods for future interannual and decadal scale climate projections.

SST↗

DOE OSTI2024

A deep learning-guided automated workflow in LipidOz for detailed characterization of fungal fatty acid unsaturation by ozonolysis

Understanding fungal lipid biology and metabolism is critical for antifungal target discovery as lipids play central roles in cellular processes. Nuances in lipid structural differences can significantly impact their functions, making it necessary to characterize lipids in detail to enable and understanding of their roles in these complex systems. In particular, lipid double bond (DB) locations are an important component of lipid structure that can only be determined using a few specialized analytical techniques. Ozone-induced dissociation mass spectrometry (OzID-MS) is one such technique that uses ozone to break lipid DBs, producing pairs of characteristic fragments that allow the determination of DB positions. In this work we apply OzID-MS and LipidOz software to analyze the complex lipids of Saccharomyces cerevisiae yeast strains transfected with different fatty acid desaturases from Histoplasma capsulatum to determine the specific unsaturated lipids produce. The automated data analysis in LipidOz made the determination of DB positions from this large dataset more practical, but manual verification for all targets was still time-consuming. The DL model reduces manual involvement in data analysis, but since it was trained using mammalian lipid extracts, the prediction accuracy on yeast-derived data was reduced. We addressed both shortcomings by retraining the DL model to act as a pre-filter to prioritize targets for automated analysis, providing confident manually verified results but requiring less computational time and manual effort. Our workflow resulted in the determination of novel DB positions and enzymatic specificity.

mass spectrometry, deep learning, Lipidomics, doub↗

DOE OSTI2024

A Dataset of CFD Simulated Industrial Furnace Images for Conditional Automatic Generation with GANs

The steel industry is constantly looking for ways to automate processes and improve efficiency. A standard practice in industry is to simulate how complex systems will operate before they are actually used. Some complex systems, including steel industry processes such as blast furnaces, require complex physics-based simulations utilizing computational fluid dynamics (CFD). These CFD physics-based simulations are very accurate but can take significant time and computational resources to process, resulting in challenges for the implementation of the models in real-world operational environments. In recent years, deep learning (DL) has been considered as a substitute for these CFD models. DL models can be trained on validated CFD simulation data and then used for industrial process inference. Previous DL-based solutions have made great contributions for industrial automation but are currently missing the additional visualization component that CFD simulations also provide. In this paper, we propose a dataset for simple DL generative approaches that can help to address this issue. The dataset and methodology under development to approach this prediction are discussed in this work.

Calix, Ricardo↗

DOE OSTI2021

Deep learning inversion of gravity data for detection of CO 2 plumes in overlying aquifers

In this work, we developed an effective U-Net based deep learning (DL) model for inversion of surface gravity data on a rectangular grid to predict 2-D high-resolution subsurface CO 2 distribution along a vertical cross-section due to CO 2 leakage through a wellbore within a deep CO 2 storage reservoir. We used synthetic data to model two types of CO 2 leakage scenarios: one CO 2 plume in a shallow aquifer (single plume case), and two plumes present at different depths (double plume case). The 3-D synthetic plume samples were created by sampling among predetermined CO 2 plume depths, saturations, and volumes. The corresponding surface gravity data on a rectangular grid were generated by a 3-D forward model. The U-Net model detected 72% of single-plume samples, and one or both plumes in 75% of double-plume samples. Most of the undetected single plumes have small gravity field strengths below the typical noise level of 5 μGal. This model generated reproducible, reliable predictions with acceptable errors and demonstrated improved spatial resolution over the conventional least-squares inversion. In contrast to the conventional least-squares inversion, which often overestimates the size of its target and underestimates its density, this U-Net model accurately delineated the boundary of a target. Furthermore, this DL inversion detected deep, small, or low saturation CO 2 plumes that are often more difficult to resolve with conventional gravity inversion methods. We note the limitations of this feasibility study, including the use of synthetic data with regular CO 2 plume shapes, and the prediction of a 2-D plume cross-section rather than the full 3-D plume, as well, we recognize the lower detection fraction for double-plume scenarios. Nevertheless, this study demonstrates that DL gravity inversion is a promising and potentially superior method to conventional least-squares inversion. Our U-Net based deep learning inversion approach may be adapted for inversion of other types of geophysical data. DL inversion can facilitate near real-time monitoring of geologic carbon sequestration to provide site operators with prompt information about subsurface CO 2 distribution for risk management and mitigation.

58 GEOSCIENCES↗

DOE OSTI2022

Towards multi-fidelity deep learning of wind turbine wakes

We report engineering wake models that accurately predict wake in a computationally efficient manner are very important for tasks such as layout optimization and control of wind farms. In this paper, we explore an application of deep learning (DL) to learn the wake model from hierarchies of physics-based approaches ranging from analytical models to an approximate form of the Reynolds-averaged Navier-Stokes equations. We first illustrate the application of principal component analysis to obtain a lower-dimensional representation that allows a computationally tractable training and deployment of DL models. Then, the DL model is trained to learn the mapping from input parameter space to the principal components, which are then used to reconstruct the three-dimensional flow field. Additionally, we investigate a composite framework consisting of two neural networks to learn the correlation between low- and high-fidelity data with Gauss and curl models treated as proxies for low- and high-fidelity models, respectively. The prediction from both DL models matches well with the high-fidelity data with a maximum relative percentage error for the kinetic energy flux of <1%. This work opens up possibilities for data-efficient construction of surrogate models for wake prediction that can be used to study the influence of wind speed and yaw angles on wind farm power production.

17 WIND ENERGY↗

DOE OSTI2022

The Data Synergy Effects of Time-Series Deep Learning Models in Hydrology

When fitting statistical models to variables in geoscientific disciplines such as hydrology, it is a customary practice to stratify a large domain into multiple regions (or regimes) and study each region separately. Traditional wisdom suggests that models built for each region separately will have higher performance because of homogeneity within each region. However, each stratified model has access to fewer and less diverse data points. Here, through two hydrologic examples (soil moisture and streamflow), we show that conventional wisdom may no longer hold in the era of big data and deep learning (DL). We systematically examined an effect we call data synergy, where the results of the DL models improved when data were pooled together from characteristically different regions. The performance of the DL models benefited from modest diversity in the training data compared to a homogeneous training set, even with similar data quantity. Moreover, allowing heterogeneous training data makes eligible much larger training datasets, which is an inherent advantage of DL. A large, diverse data set is advantageous in terms of representing extreme events and future scenarios, which has strong implications for climate change impact assessment. The results here suggest the research community should place greater emphasis on data sharing.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Use of Satellite, Surface Observations and Numerical Weather Prediction Model Data to Improve Cloud Base Height and Cloud Base Vertical Velocity Estimation

Cloud base height (CBH) and cloud base vertical velocity (CBVV) are important variables that impact the overall climate in a region as they influence the formulation, longevity, and evolution of clouds. Retrieval of both parameters have long used ground instrumentation (e.g., Doppler lidar (DL), ground base radar); however, retrieving CBH from satellites is particularly challenging given that space-based instruments only observe cloud tops. In this manuscript, CBH is retrieved using a multi-linear regression equation, while CBVV used a random forests model. Both retrievals combine satellite and numerical weather prediction data. The satellite data used are the Visible Infrared Imaging Radiometer Suite imagery, while measurements of CBH and CBVV include DL and radiosonde data at the Southern Great Plains (SGP) Atmospheric Radiation Measurement observatory. Data from 83 summer days (May-August) in 2018–2021 featuring cumulus clouds forced by solar heating were examined and used to train the models, with years 2022–2023 used for validation. Various spatial domains were defined with one large (2.4° longitude by 2.0° latitude) SGP domain being split into smaller sections (smallest being 0.99° and 0.61° longitude and latitude respectably). CBH and CBVV values obtained from the DL as compared to the models show root mean square errors between 150 and 200 m, with CBVV values between 0.45 and 1 ms -1 . Finally, it was found that the CBH formulation performs well over all domains, while the CBVV retrievals become less accurate due to more turbulence being introduced into the observations as the number of DL stations decreases in the smaller domains.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2023

Explainable deep learning for insights in El Niño and river flows

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2022

Deep learning-based noise filtering toward millisecond order imaging by using scanning transmission electron microscopy

Application of scanning transmission electron microscopy (STEM) to in situ observation will be essential in the current and emerging data-driven materials science by taking STEM’s high affinity with various analytical options into account. As is well known, STEM’s image acquisition time needs to be further shortened to capture a targeted phenomenon in real-time as STEM’s current temporal resolution is far below the conventional TEM’s. However, rapid image acquisition in the millisecond per frame or faster generally causes image distortion, poor electron signals, and unidirectional blurring, which are obstacles for realizing video-rate STEM observation. Here we show an image correction framework integrating deep learning (DL)-based denoising and image distortion correction schemes optimized for STEM rapid image acquisition. By comparing a series of distortion corrected rapid scan images with corresponding regular scan speed images, the trained DL network is shown to remove not only the statistical noise but also the unidirectional blurring. This result demonstrates that rapid as well as high-quality image acquisition by STEM without hardware modification can be established by the DL. The DL-based noise filter could be applied to in-situ observation, such as dislocation activities under external stimuli, with high spatio-temporal resolution.

47 OTHER INSTRUMENTATION↗

DOE OSTI2022

Reduced order modeling for flow and transport problems with Barlow Twins self-supervised learning

Abstract We propose a unified data-driven reduced order model (ROM) that bridges the performance gap between linear and nonlinear manifold approaches. Deep learning ROM (DL-ROM) using deep-convolutional autoencoders (DC–AE) has been shown to capture nonlinear solution manifolds but fails to perform adequately when linear subspace approaches such as proper orthogonal decomposition (POD) would be optimal. Besides, most DL-ROM models rely on convolutional layers, which might limit its application to only a structured mesh. The proposed framework in this study relies on the combination of an autoencoder (AE) and Barlow Twins (BT) self-supervised learning, where BT maximizes the information content of the embedding with the latent space through a joint embedding architecture. Through a series of benchmark problems of natural convection in porous media, BT–AE performs better than the previous DL-ROM framework by providing comparable results to POD-based approaches for problems where the solution lies within a linear subspace as well as DL-ROM autoencoder-based techniques where the solution lies on a nonlinear manifold; consequently, bridges the gap between linear and nonlinear reduced manifolds. We illustrate that a proficient construction of the latent space is key to achieving these results, enabling us to map these latent spaces using regression models. The proposed framework achieves a relative error of 2% on average and 12% in the worst-case scenario (i.e., the training data is small, but the parameter space is large.). We also show that our framework provides a speed-up of $$7 \times 10^{6}$$ 7 × 10 6 times, in the best case, and $$7 \times 10^{3}$$ 7 × 10 3 times on average compared to a finite element solver. Furthermore, this BT–AE framework can operate on unstructured meshes, which provides flexibility in its application to standard numerical solvers, on-site measurements, experimental data, or a combination of these sources.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Self-consistent theory of the Darrieus–Landau and Rayleigh–Taylor instabilities with self-generated magnetic fields

The Rayleigh–Taylor (RT) and Darrieus–Landau (DL) instabilities are studied in an inertial confinement fusion context within the framework of small critical-to-shell density ratio D R and weak acceleration regime, i.e., large Froude number Fr. The quasi-isobaric analysis in Sanz et al. is completed with the inclusion of non-isobaric and self-generated magnetic-field effects. The analysis is restricted to perturbation wavelengths k~1 larger than the conduction length scale at the ablation front, yet its validity ranges from wavelengths shorter and larger than the conduction layer width (distance between the ablation front and the critical surface). The use of a sharp boundary model leads to a single analytical expression of the dispersion relation encompassing both instabilities. The two new effects come into play by modifying the perturbed mass and momentum fluxes at the ablation front. The momentum flux (perturbed pressure at the spike) is the predominant stabilizing mechanism in the RT instability (overpressure) and the driving mechanism in the DL instability (underpressure). The non-isobaric effects notably modify the scaling laws in the DL limit, leading to an underpressure scaling as ~k –11/15 rather than ~k –2/5 obtained in the quasi-isobaric model. The magnetic fields are generated due to misalignment between pressure and density gradients (Biermann battery effect). They affect the hydrodynamics by bending the heat flux lines. Within the framework of this paper, they enhance ablation, resulting in a stabilizing effect that peaks for perturbation wavelengths comparable to the conduction layer width. Here, the combination of parameters D R Fr 2/3 defines the region of predominance of each instability in the dispersion relation. It is proven that the DL region falls outside of the parameter range in inertial confinement fusion.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2025

Evaluating probabilistic deep learning methods for uncertainty quantification of temperature downscaling

Deep learning (DL) has emerged as a promising tool for downscaling coarse-resolution climate data to high-resolution outputs, enabling improved regional climate predictions. A critical aspect of DL-based downscaling is the incorporation of uncertainty quantification (UQ), which enhances the interpretability and reliability of predictions—key factors for climate risk assessment and decision-making. This study develops a DL model to downscale 2 m temperature across the contiguous United States using reanalysis datasets. We systematically evaluate three epistemic UQ methods—deep ensembles (DEns), Monte Carlo dropout (MCD), and Flipout—based on their probabilistic accuracy, downscaling performance, sensitivity to geographical features, and computational efficiency. Results indicate that MCD generally outperforms Flipout and DEns in terms of calibration and downscaling accuracy. However, DEns demonstrate lower calibration errors in coastal regions, indicating its higher confidence within these areas. Flipout, in contrast, is more sensitive to elevation gradients and exhibits higher calibration errors in mountainous regions. Hence, the choice of UQ method for this task depends on the specific requirements of the application. For applications that prioritize overall calibration, downscaling accuracy, and computational efficiency, MCD is a strong candidate. These findings highlight the importance of selecting UQ methods based on application-specific requirements, such as geographical context and computational constraints. By addressing the trade-offs between UQ methods, this study provides actionable insights for improving the reliability, scalability, and utility of DL-based downscaling in climate science.

Environmental sciences↗

DOE OSTI2023

Deep Learning Approach for High-accuracy Electron Counting of Monolithic Active Pixel Sensor-type Direct Electron Detectors at Increased Electron Dose

Abstract Electron counting can be performed algorithmically for monolithic active pixel sensor direct electron detectors to eliminate readout noise and Landau noise arising from the variability in the amount of deposited energy for each electron. Errors in existing counting algorithms include mistakenly counting a multielectron strike as a single electron event, and inaccurately locating the incident position of the electron due to lateral spread of deposited energy and dark noise. Here, we report a supervised deep learning (DL) approach based on Faster region-based convolutional neural network (R-CNN) to recognize single electron events at varying electron doses and voltages. The DL approach shows high accuracy according to the near-ideal modulation transfer function (MTF) and detector quantum efficiency for sparse images. It predicts, on average, 0.47 pixel deviation from the incident positions for 200 kV electrons versus 0.59 pixel using the conventional counting method. The DL approach also shows better robustness against coincidence loss as the electron dose increases, maintaining the MTF at half Nyquist frequency above 0.83 as the electron density increases to 0.06 e−/pixel. Thus, the DL model extends the advantages of counting analysis to higher dose rates than conventional methods.

Materials Science↗

DOE OSTI2023

How the birth and death of shear layers determine confinement evolution: from the L → H transition to the density limit

Electric field profile structure—especially its shear—is a natural order parameter for the edge plasma, and characterizes confinement regimes ranging from the H-mode (Wagner et al. 1982 Phys. Rev. Lett.49, 1408–1412 (doi:10.1103/PhysRevLett.49.1408)) to the density limit (DL) (Greenwald et al. 1988 Nucl. Fusion28, 2199–2207 (doi:10.1088/0029-5515/28/12/009)). The theoretical developments and lessons learned during 40 years of H-mode studies (Connor & Wilson 1999 Plasma Phys. Control. Fusion42, R1–R74 (doi:10.1088/0741-3335/42/1/201); Wagner 2007 Plasma Phys. Control. Fusion49, B1–B33 (doi:10.1088/0741-3335/49/12b/s01)) are applied to the shear layer collapse paradigm (Hong et al. 2017 Nucl. Fusion58, 016041 (doi:10.1088/1741-4326/aa9626)) for the onset of DL phenomena. Results from recent experiments on edge shear layers and DL phenomenology are summarized and discussed in the light of L → H transition physics. The theory of shear layer collapse is then developed. In this work, we demonstrate that shear layer physics captures both the well known current (Greenwald) scaling of the DL (Greenwald 2002 Plasma Phys. Control. Fusion44, R27–R53 (doi:10.1088/0741-3335/44/8/201); Greenwald et al. 2014 Phys. Plasmas21, 110501 (doi:10.1063/1.4901920)), as well as the emerging power scaling (Zanca, Sattin, JET Contributors 2019 Nucl. Fusion59, 126011 (doi:10.1088/1741-4326/ab3b31)). The derivation of the power scaling theory exploits an existing model, originally developed for the L → H transition (Diamond, Liang, Carreras, Terry 1994 Phys. Rev. Lett.72, 2565–2568 (doi:10.1103/PhysRevLett.72.2565); Kim & Diamond 2003 Phys. Rev. Lett.90, 185006 (doi:10.1103/PhysRevLett.90.185006)). We describe the enhanced particle transport events that occur following shear layer collapse. Open problems and future directions are discussed.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2022

Hvac: Removing I/O Bottleneck for Large-Scale Deep Learning Applications

Scientific communities are increasingly adopting deep learning (DL) models in their applications to accelerate scientific discovery processes. However, with rapid growth in the computing capabilities of HPC supercomputers, large-scale DL applications have to spend a significant portion of training time performing I/O to a parallel storage system. Previous research works have investigated optimization techniques such as prefetching and caching. Unfortunately, there exist non-trivial challenges to adopting the existing solutions on HPC supercomputers for large-scale DL training applications, which include non-performance and/or failures at extreme scale, lack of portability and generality in design, complex deployment methodology, and being limited to a specific application or dataset. To address these challenges, we propose High-Velocity AI Cache (HVAC), a distributed read-cache layer that targets and fully exploits the node-local storage or near node-local storage technology. HVAC seamlessly accelerates read I/O by aggregating node-local or near node-local storage, avoiding metadata lookups and file locking while preserving portability in the application code. We deploy and evaluate HVAC on 1,024 nodes (with over 6000 NVIDIA V100 GPUS) of the Summit supercomputer. In particular, we evaluate the scalability, efficiency, accuracy, and load distribution of HVAC compared to GPFS and XFS-on-NVMe. With four different DL applications, we observe an average 25 % performance improvement atop GPFS and 9% drop against XFS-on-NVMe, which scale linearly and are considered the performance upper bound. We envision HVAC as an important caching library for upcoming HPC supercomputers such as Frontier.

Khan, Awais↗

DOE OSTI2025

Advancing the Frontiers of Deep Learning for Low-Dose 3D Cone-Beam CT Reconstruction

X-ray computed tomography (CT) is an important noninvasive medical imaging modality for studying the structural details of internal organs. Image reconstruction in CT is an inverse problem of recovering an object's internal structure from the absorption profile of X-ray beams (sinogram) measured using a detector. The classical variational approach for CT reconstruction minimizes an energy functional using an appropriate iterative algorithm. Motivated by the success of deep learning (DL), researchers have begun to leverage training data and enhanced computing capabilities in recent years to produce high-fidelity reconstructed images. Nonetheless, much of the academic research in DL algorithms for CT has focused primarily on the two-dimensional setting (with simplified forward operators and noise model) for proofs-of-concept, and a comprehensive benchmarking of various classical and data-driven CT reconstruction approaches has not beenundertaken. The key objective of our CT reconstruction grand challenge was to promote methodological advancements for both classical and DL-based approaches for clinical CT with a reasonably accurately simulated 3D CT forward operator and noise model. We have utilized the publicly available LIDC-IDRI dataset and simulated sinograms and FDK images corresponding to two dose levels (clinical- and low-dose, constituting two tracks of the challenge) starting from the normal-dose images as the ground truth. In this paper, we summarize the motivation, context, and results of our challenge, and highlight the future research directions in DL for clinical CT.

X-ray tomography↗

DOE OSTI2023

Reducing Urban Traffic Congestion Using Deep Learning and Model Predictive Control

This article proposes a deep learning (DL)-based control algorithm—DL velocity-based model predictive control (VMPC)—for reducing traffic congestion with slowly time-varying traffic signal controls. This control algorithm consists of system identification using DL and traffic signal control using VMPC. For the training process of DL, we established a modeling error entropy loss as the criteria inspired by the theory of stochastic distribution control (SDC) originated by the fourth author. Herein, simulation results show that the proposed algorithm can reduce traffic congestion with a slowly varying traffic signal control input. Results of an ablation study demonstrate that this algorithm compares favorably to other model-based controllers in terms of prediction error, signal varying speed, and control effectiveness.

97 MATHEMATICS AND COMPUTING↗