Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “empirical deep learning”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

BUTTER - Empirical Deep Learning Dataset

The BUTTER Empirical Deep Learning Dataset represents an empirical study of the deep learning phenomena on dense fully connected networks, scanning across thirteen datasets, eight network shapes, fourteen depths, twenty-three network sizes (number of trainable parameters), four learning rates, six minibatch sizes, four levels of label noise, and fourteen levels of L1 and L2 regularization each. Multiple repetitions (typically 30, sometimes 10) of each combination of hyperparameters were preformed, and statistics including training and test loss (using a 80% / 20% shuffled train-test split) are recorded at the end of each training epoch. In total, this dataset covers 178 thousand distinct hyperparameter settings ("experiments"), 3.55 million individual training runs (an average of 20 repetitions of each experiments), and a total of 13.3 billion training epochs (three thousand epochs were covered by most runs). Accumulating this dataset consumed 5,448.4 CPU core-years, 17.8 GPU-years, and 111.2 node-years.

Array↗
DOE OSTI2022

BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset

The BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset adds node-level energy consumption data from watt-meters to the primary sweep of the BUTTER - Empirical Deep Learning Dataset. This dataset contains energy consumption and performance data from 63,527 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (number of trainable parameters), 8 network "shapes", and 14 depths on both CPU and GPU hardware collected using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use, and highlights the impact of cache effects. BUTTER-E is intended to be joined with the BUTTER dataset (see "BUTTER - Empirical Deep Learning Dataset on OEDI" resource below) which characterizes the performance of 483k distinct fully connected neural networks but does not include energy measurements.

Array↗
DOE OSTI2026

Measuring the Energy Consumption and Efficiency of Deep Neural Networks: An Empirical Analysis and Design Recommendations

Addressing the "Red-AI" trend of rising energy consumption by large-scale neural networks, this study investigates the measured energy consumption of training various fully connected neural network architectures. We introduce the BUTTER-E dataset, an augmentation to the BUTTER Empirical Deep Learning dataset, containing energy consumption and performance data from 41,129 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (trainable parameters), 8 "shapes", and 14 depths on both CPUs and GPUs using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use. Our analysis uncovers a surprising, hardware-mediated non-linear relationship between energy efficiency and network design, challenging the assumption that reducing the number of parameters or FLOPs is the best way to achieve greater energy efficiency. We propose a straightforward and effective energy model that accounts for network size, computing, and memory hierarchy. Highlighting the need for cache-considerate algorithm development, we suggest a codesign approach to energy efficient network, algorithm, and hardware design. This work contributes to the fields of sustainable computing and Green AI, offering practical guidance for creating more energy-efficient neural networks and promoting sustainable AI.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Green Computing Catalyzer Kickoff: Green Artificial Intelligence (AI) Directions

For the 2022 JISEA annual meeting, JISEA introduces in this presentation the new Green Computing Catalyzer. The new catalyzer looks to establish green computing as a salient research domain at NREL, and crucially, to begin to address the looming computing energy crisis. The presentation provides some basic numbers, establishing energy use and source or carbon emissions for computing. The presentation also includes a brief history of green computing at NREL, empirical deep learning efficiency study, and research directions.

AI efficiency↗
DOE OSTI2025

Deep-learning atomistic semi-empirical pseudopotential model for nanomaterials

The semi-empirical pseudopotential method (SEPM) has been widely applied to provide computational insights into the electronic structure, photophysics, and charge carrier dynamics of nanoscale materials. We present “DeepPseudopot”, a machine-learned atomistic pseudopotential model that extends the SEPM framework by combining a flexible neural network representation of the local pseudopotential with parameterized non-local and spin-orbit coupling terms. Trained on bulk quasiparticle band structures and deformation potentials from GW calculations, the model captures many-body and relativistic effects with very high accuracy across diverse semiconducting materials, as illustrated for silicon and group III-V semiconductors. DeepPseudopot’s accuracy, efficiency, and transferability make it well-suited for data-driven in silico design and discovery of novel optoelectronic nanomaterials.

Lin, Kailai [University of California, Berkeley, C↗
DOE OSTI2022

Green AI: Insights Into Deep Learning's Looming Energy Efficiency Crisis

As demands grow to integrate artificial intelligence into every aspect of industry, commerce, and life, deep learning's exploding energy cost has become a looming crisis, making AI systems a salient energy-efficiency challenge. One might expect that doubling a neural network's size would halve its error rate, or at least allow it to achieve greater performance given the same amount of time and energy. I will present clear and substantial scientific evidence which indicates that not only is this intuition wildly wrong, but that neural networks scale so poorly that to increase deep learning performance by only a small fraction can easily require an order of magnitude or more increase in computational resources and energy. Further, the marginal trade-off price of to increase model performance rapidly explodes as performance targets are increased. To address this challenge, I will provide a toolkit of techniques that can be applied today to mitigate the inefficiency of modern deep learning. And, I will conclude by illuminating a practical path forward towards efficient, Green AI.

artificial intelligence↗
DOE OSTI2023

Finite-Time Analysis of Whittle Index based Q-Learning for Restless Multi-Armed Bandits with Neural Network Function Approximation

Whittle index policy is a heuristic to the intractable restless multi-armed bandits (RMAB) problem. Although it is provably asymptotically optimal, finding Whittle indices remains difficult. In this paper, we present Neural-Q-Whittle, a Whittle index based Q-learning algorithm for RMAB with neural network function approximation, which is an example of nonlinear two-timescale stochastic approximation with Q-function values updated on a faster timescale and Whittle indices on a slower timescale. Despite the empirical success of deep Q-learning, the non-asymptotic convergence rate of Neural-Q-Whittle, which couples neural networks with two-timescale Q-learning largely remains unclear. This paper provides a finite-time analysis of Neural-Q-Whittle, where data are generated from a Markov chain, and Q-function is approximated by a ReLU neural network. Our analysis leverages a Lyapunov drift approach to capture the evolution of two coupled parameters, and the nonlinearity in value function approximation further requires us to characterize the approximation error. Combing these provide Neural-Q-Whittle with convergence rate, where is the number of iterations.

reinforcement learning, structured learning, conve↗
DOE OSTI2025

Dataset, Code, and Models for Training Deep Learning Potentials for Low Temperature Plasma-Surface Interactions

This repository contains datasets, training scripts, and finished models, and test simulations used in the development of DeepREBO— a machine-learned interatomic potential trained to emulate the REBO2 empirical potential. The data was generated to study deep potential development for simulations of plasma-surface interactions. It uses an active learning framework, starting from a minimal dataset and iteratively expanding it. Included are those generated datasets, the trained models, and simulations used to evaluate the performance of the training process. This resource supports reproducibility and provides a reference framework for training deep potentials in plasma-surface interaction studies.

active learning↗
DOE OSTI2025

Development and transferability of neural-network models for plasma-surface interactions

Plasma-surface interactions are increasingly critical to modern technologies; yet, accurate molecular dynamics simulations remain limited by the capabilities of interatomic potentials. Deep Potentials (DPs) promise to revolutionize the field by providing a systematic method for producing accurate interatomic potentials. The primary challenge of DP development is selecting a dataset, which efficiently spans the set of atomic environments one expects to encounter in the subsequent molecular dynamics simulations. The computational cost of density functional theory calculations, which are the typical basis for DP development, makes it impossible to directly verify the quality of a given DP. To address this challenge, we explore the development of a deep-learned interatomic potential, “DeepREBO,” trained to reproduce the behavior of the REBO2 empirical potential, enabling direct validation of training methodology and transferability. Using an active learning framework, we begin with a minimal dataset and iteratively expand it to train a Deep Potential-Smooth Edition model that faithfully reproduces REBO2 results for 25 eV hydrogen bombardment of diamond (001), a particularly challenging case. We show that small, carefully curated datasets can outperform large, unguided ones, with effective models requiring fewer than 15 000 snapshots. Subsequent transferability tests demonstrate that while DeepREBO generalizes well to diamond (111) surfaces, performance degrades for amorphous carbon or higher-energy impacts, highlighting the need for use-case-specific training data. We also evaluate methods to improve short-range repulsion. This study outlines best practices for training robust deep potentials and underscores the importance of dataset design for predictive plasma simulations.

Ab-initio molecular dynamics↗
DOE OSTI2025

Testing convolutional neural network based deep learning systems: a statistical metamorphic approach

Machine learning technology spans many areas and today plays a significant role in addressing a wide range of problems in critical domains,i.e., healthcare, autonomous driving, finance, manufacturing, cybersecurity,etc. Metamorphic testing (MT) is considered a simple but very powerful approach in testing such computationally complex systems for which either an oracle is not available or is available but difficult to apply. Conventional metamorphic testing techniques have certain limitations in verifying deep learning-based models (i.e., convolutional neural networks (CNNs)) that have a stochastic nature (because of randomly initializing the network weights) in their training. In this article, we attempt to address this problem by using a statistical metamorphic testing (SMT) technique that does not require software testers to worry about fixing the random seeds (to get deterministic results) to verify the metamorphic relations (MRs). We propose seven MRs combined with different statistical methods to statistically verify whether the program under test adheres to the relation(s) specified in the MR(s). We further use mutation testing techniques to show the usefulness of the proposed approach in the healthcare space and test two CNN-based deep learning models (used for pneumonia detection among patients). The empirical results show that our proposed approach uncovers 85.71% of the implementation faults in the classifiers under test (CUT). Furthermore, we also propose an MRs minimization algorithm for the CUT, thus saving computational costs and organizational testing resources.

Computer Science↗
DOE OSTI2022

Scalable deep learning for watershed model calibration

Watershed models such as the Soil and Water Assessment Tool (SWAT) consist of high-dimensional physical and empirical parameters. These parameters often need to be estimated/calibrated through inverse modeling to produce reliable predictions on hydrological fluxes and states. Existing parameter estimation methods can be time consuming, inefficient, and computationally expensive for high-dimensional problems. In this paper, we present an accurate and robust method to calibrate the SWAT model (i.e., 20 parameters) using scalable deep learning (DL). We developed inverse models based on convolutional neural networks (CNN) to assimilate observed streamflow data and estimate the SWAT model parameters. Scalable hyperparameter tuning is performed using high-performance computing resources to identify the top 50 optimal neural network architectures. We used ensemble SWAT simulations to train, validate, and test the CNN models. We estimated the parameters of the SWAT model using observed streamflow data and assessed the impact of measurement errors on SWAT model calibration. We tested and validated the proposed scalable DL methodology on the American River Watershed, located in the Pacific Northwest-based Yakima River basin. Our results show that the CNN-based calibration is better than two popular parameter estimation methods (i.e., the generalized likelihood uncertainty estimation [GLUE] and the dynamically dimensioned search [DDS], which is a global optimization algorithm). For the set of parameters that are sensitive to the observations, our proposed method yields narrower ranges than the GLUE method but broader ranges than values produced using the DDS method within the sampling range even under high relative observational errors. The SWAT model calibration performance using the CNNs, GLUE, and DDS methods are compared using R 2 and a set of efficiency metrics, including Nash-Sutcliffe, logarithmic Nash-Sutcliffe, Kling-Gupta, modified Kling-Gupta, and non-parametric Kling-Gupta scores, computed on the observed and simulated watershed responses. The best CNN-based calibrated set has scores of 0.71, 0.75, 0.85, 0.85, 0.86, and 0.91. The best DDS-based calibrated set has scores of 0.62, 0.69, 0.8, 0.77, 0.79, and 0.82. The best GLUE-based calibrated set has scores of 0.56, 0.58, 0.71, 0.7, 0.71, and 0.8. The scores above show that the CNN-based calibration leads to more accurate low and high streamflow predictions than the GLUE and DDS sets. Our research demonstrates that the proposed method has high potential to improve our current practice in calibrating large-scale integrated hydrologic models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Sparse Data Machine Learning Integration with Theory, Experiment and Uncertainty Quantification: Process-Structure-Property-Performance of Friction Deformation Processing

Computer vision and deep learning tools that advance the ability to establish processing-structure-property-performance (PSPP) relations are presented. The Bayesian binning method for image segmentation enables quantitative analysis of microstructural features in an automated way, while the analysis of shapes and relative orientation of these features reveals local deformation maps indicative of both, material flow and residual stresses due to materials processing. The deep learning method leads to the previous knowledge agnostic mapping of empirically observed microstructural zones in friction stir welding (FSW) process and synthetic microstructure generation capability that is statistically equivalent to experimentally collected data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Deep potential molecular dynamics simulations of low-temperature plasma-surface interactions

Machine learning approaches to potential generation for molecular dynamics (MD) simulations of low-temperature plasma-surface interactions could greatly extend the range of chemical systems that can be modeled. Empirical potentials are difficult to generalize to complex combinations of multiple elements with interactions that might include covalent, ionic, and metallic bonds. This work demonstrates that a specific machine learning approach, Deep Potential Molecular Dynamics (DeepMD), can generate potentials that provide a good model of plasma etching in the Si-Cl-Ar system. Comparisons are made between MD results using DeepMD models and empirical potentials, as well as experimental measurements. Pure Si properties predicted by the DeepMD model are in reasonable agreement with experimental results. Simulations of Si bombardment by Ar + ions demonstrate the ability of the DeepMD method to predict sputtering yields as well as the depth of the amorphous-crystalline interface. Etch yields as a function of flux ratio and ion energy for simultaneous Cl 2 and Ar + impacts are in good agreement with previous simulation results and experiment. Predictions of etch yields and etch products during plasma-assisted atomic layer etching of Si-Cl 2 -Ar are shown to be in good agreement with MD predictions using empirical potentials and with experiment. Finally, good agreement was also seen with measurements for the spontaneous etching of Si by Cl atoms at 300 K. Further, the demonstration that DeepMD can reproduce results from MD simulations using empirical potentials is a necessary condition to future efforts to extend the method to a much wider range of systems for which empirical potentials may be difficult or impossible to obtain.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRSDate not supplied

Generation of Continental Scale Percent Tree Cover Product Using Deep-learning and Multi-scale Remote Sensing Data

Spatially explicit percent tree cover (TC) estimation is critical for mapping forest aboveground biomass and its dynamics. While various TC products have been developed, there has not been a generalized framework that can be applied to diverse terrestrial ecosystems due to underlain extreme complexities. Deep learning algorithms can learn a spatial pattern and radiometric characteristics of tree canopy as a robust approximation of physical or empirical models, and thus have emerged as promising and efficient tools for large-scale TC mapping. In this study, we synergistically use very high-resolution aerial imageries (National Agriculture Imagery Program, NAIP) and medium resolution Landsat data to map continental-scale TC (CONUS and Mexico) through a hierarchical deep learning approach (Convolutional Neural Network), i.e., NAIP TC generated from a NAIP model is utilized to train a Landsat model. The produced TC product (hereafter, NEX-TC) is able to capture the spatial pattern of TC distribution and its changes driven by natural disturbance and human land management. We further explore and analyze the reliability and potential uncertainty of the NEX-TC by comparing it to lidar- (lidar-TC), National Land Cover Database (NLCD-TC), and MODIS Vegetation Continuous Field (MODIS-TC). This evaluation practice reveals that TC products based on passive optical sensors tend to underestimate TC across all land cover types while Landsat-based TCs (i.e., NEX-TC & NLCD-TC) perform better than the coarser MODIS TC estimate. Our results show that the NEX-TC is generally comparable to NLCD-TC but it particularly outperforms NLCD-TC and MODIS-TC over the dense forests where lidar-TC indicates >80% TC. These results indicate that our hierarchical deep learning approach and TC product will be effective and useful for characterizing large-scale tree cover and possibly associated carbon dynamics.

Landsat↗
DOE OSTI2025

Toward a Generalizable Prediction Model of Molten Salt Mixture Density with Chemistry‐Informed Transfer Learning

Optimally designing applications of molten salts requires knowledge of their thermophysical properties over a wide range of temperatures and compositions. There exist significant gaps in existing databases and this data can be challenging to experimentally measure due to high temperatures, salt corrosivity, and salt hygroscopicity. Existing databases have been used to create Redlich–Kister (RK) models for mixture density showing improved accuracy with respect to ideal mixing assumptions, but these models require subcomponent data measurements for each new system, therefore lacking generality. In order to address generalizability and data sparsity, a transfer learning procedure is proposed to train deep neural networks (DNNs) using a combination of semi‐empirical relationships (RK), data from the thermophysical arm of the molten salt thermal properties database and universal ab initio properties of component mixtures taken from the joint automated repository for various integrated simulations (JARVIS) classical force‐field inspired descriptors database to predict density in molten salts. Herein, it is shown that DNNs predict molten salt density with an r 2 over 0.99 and a mean absolute percentage error under 1%, outperforming alternative methods.

inorganic materials↗
DOE OSTI2023

Deep neural network uncertainty quantification for LArTPC reconstruction

We evaluate uncertainty quantification (UQ) methods for deep learning applied to liquid argon time projection chamber (LArTPC) physics analysis tasks. As deep learning applications enter widespread usage among physics data analysis, neural networks with reliable estimates of prediction uncertainty and robust performance against overconfidence and out-of-distribution (OOD) samples are critical for their full deployment in analyzing experimental data. While numerous UQ methods have been tested on simple datasets, performance evaluations for more complex tasks and datasets are scarce. Here we assess the application of selected deep learning UQ methods on the task of particle classification using the PiLArNet monte carlo 3D LArTPC point cloud dataset. We observe that UQ methods not only allow for better rejection of prediction mistakes and OOD detection, but also generally achieve higher overall accuracy across different task settings. We assess the precision of uncertainty quantification using different evaluation metrics, such as distributional separation of prediction entropy across correctly and incorrectly identified samples, receiver operating characteristic curves (ROCs), and expected calibration error from observed empirical accuracy. We conclude that ensembling methods can obtain well calibrated classification probabilities and generally perform better than other existing methods in deep learning UQ literature.

47 OTHER INSTRUMENTATION↗
DOE OSTI2025

Uncertainty Quantification for Data-Driven Machine Learning Models in Nuclear Engineering Applications: Where We Are and What Do We Need?

Machine learning (ML) has been leveraged to tackle a diverse range of tasks in almost all branches of nuclear engineering. Many of the successes in ML applications can be attributed to the recent performance breakthroughs in deep learning, the growing availability of computational power, data, and easy-to-use ML libraries. However, these empirical successes have often outpaced our formal understanding of the ML algorithms. An important but under-rated area is uncertainty quantification (UQ) of ML. ML-based models are subject to approximation uncertainty when they are used to make predictions, due to sources including but not limited to, data noise, data coverage, extrapolation, imperfect model architecture and the stochastic training process. The goal of this paper is to clearly explain and illustrate the importance of UQ of ML. We will elucidate the differences in the basic concepts of UQ of physics-based models and data-driven ML models. Various sources of uncertainties in physical modeling and data-driven modeling will be discussed, demonstrated, and compared. We will also present and demonstrate a few techniques to quantify the ML prediction uncertainties, including Monte Carlo dropout, deep ensemble, Bayesian neural networks, Gaussian Processes and conformal prediction. Lastly, we will discuss the need for building a verification, validation and UQ framework to establish ML credibility.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗