Search NASA⌕ Search

SEARCH · Search NASA

Results for “Training Principles”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events

Abstract Machine learned force fields typically require manual construction of training sets consisting of thousands of first principles calculations, which can result in low training efficiency and unpredictable errors when applied to structures not represented in the training set of the model. This severely limits the practical application of these models in systems with dynamics governed by important rare events, such as chemical reactions and diffusion. We present an adaptive Bayesian inference method for automating the training of interpretable, low-dimensional, and multi-element interatomic force fields using structures drawn on the fly from molecular dynamics simulations. Within an active learning framework, the internal uncertainty of a Gaussian process regression model is used to decide whether to accept the model prediction or to perform a first principles calculation to augment the training set of the model. The method is applied to a range of single- and multi-element systems and shown to achieve a favorable balance of accuracy and computational efficiency, while requiring a minimal amount of ab initio training data. We provide a fully open-source implementation of our method, as well as a procedure to map trained models to computationally efficient tabulated force fields.

36 MATERIALS SCIENCE↗
DOE OSTI2026

A variational framework for residual-based adaptivity in neural PDE solvers and operator learning

Residual-based adaptive strategies are widely used in scientific machine learning yet remain largely heuristic. We introduce a variational framework that formalizes these methods through convex transformations of the residual, where different transformations correspond to distinct objective functionals. For instance, exponential weights target uniform error minimization, while linear weights recover quadratic error minimization. This perspective reveals adaptive weighting as a means of selecting sampling distributions that optimize a primal objective, directly linking discretization choices to error metrics. This principled approach yields three key benefits: it enables systematic design of adaptive schemes, reduces discretization error by lowering estimator variance, and enhances learning dynamics by improving gradient signal-to-noise ratio. Extending the framework to operator learning, we demonstrate substantial performance gains across diverse optimizers and architectures. Our results provide a theoretical perspective for residual-based adaptivity and establish a foundation for principled discretization and training.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2016

An analysis of image storage systems for scalable training of deep neural networks

This study presents a principled empirical evaluation of image storage systems for training deep neural networks. We employ the Caffe deep learning framework to train neural network models for three different data sets, MNIST, CIFAR-10, and ImageNet. While training the models, we evaluate five different options to retrieve training image data: (1) PNG-formatted image files on local file system; (2) pushing pixel arrays from image files into a single HDF5 file on local file system; (3) in-memory arrays to hold the pixel arrays in Python and C++; (4) loading the training data into LevelDB, a log-structured merge tree based key-value storage; and (5) loading the training data into LMDB, a B+tree based key-value storage. The experimental results quantitatively highlight the disadvantage of using normal image files on local file systems to train deep neural networks and demonstrate reliable performance with key-value storage based storage systems. When training a model on the ImageNet dataset, the image file option was more than 17 times slower than the key-value storage option. Along with measurements on training time, this study provides in-depth analysis on the cause of performance advantages/disadvantages of each back-end to train deep neural networks. We envision the provided measurements and analysis will shed light on the optimal way to architect systems for training neural networks in a scalable manner.

Lim, Seung-Hwan↗
DOE OSTI2021

Differentiable programming for online training of a neural artificial viscosity function within a staggered grid Lagrangian hydrodynamics scheme

Lagrangian methods to solve the inviscid Euler equations produce numerical oscillations near shock waves. A common approach to reducing these oscillations is to add artificial viscosity (AV) to the discrete equations. The AV term acts as a dissipative mechanism that attenuates oscillations by smearing the shock across a finite number of computational cells. However, AV introduces several control parameters that are not determined by the underlying physical model, and hence, in practice are tuned to the characteristics of a given problem. We seek to improve the standard quadratic-linear AV form by replacing it with a learned neural function that reduces oscillations relative to exact solutions of the Euler equations, resulting in a hybrid numerical-neural hydrodynamic solver. Because AV is an artificial construct that exists solely to improve the numerical properties of a hydrodynamic code, there is no offline ‘viscosity data’ against which a neural network can be trained before inserting into a numerical simulation, thus requiring online training. We achieve this via differentiable programming, i.e. end-to-end backpropagation or adjoint solution through both the neural and differential equation code, using automatic differentiation of the hybrid code in the Julia programming language to calculate the necessary loss function gradients. A novel offline pre-training step accelerates training by initializing the neural network to the default numerical AV scheme, which can be learned rapidly by space-filling sampling over the AV input space. We find that online training over early time steps of simulation is sufficient to learn a neural AV function that reduces numerical oscillations in long-term hydrodynamic shock simulations. These results offer an early proof-of-principle that online differentiable training of hybrid numerical schemes with novel neural network components can improve certain performance aspects existing in purely numerical schemes.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Data-driven emulation of modal aerosol microphysics via neural operator-based modeling

The complexity and the small characteristic scales of aerosol microphysical processes pose a big challenge for accurate and efficient Earth system simulations at regional and global scales. In this work, we construct and evaluate a surrogate model: the aerosol deep operator network (ADON), a physics-inspired dual-net architecture for emulating the aerosol microphysics parameterization suite in the version 2 of the Energy Earth System Model (E3SMv2). The current version of the surrogate model is trained on a dataset comprising 9.8 million samples obtained from a global E3SMv2 simulation with the horizontal resolution of about one degree under cloud-free conditions. Incorporating domain spatial and temporal coordinates, as well as principle components extracted from training data, the dual-net surrogate model effectively captures the intricate representations of aerosol and the relationship with atmospheric state variables, achieving an R-squared score over $$95.7\%$$ for all the lognormal aerosol modes in the extrapolated regime. The validated model provides feature importance of input variables and their impact on the predictive capacity of the surrogate model in relation to the E3SM. The computational cost of online inference time deployed on CPUs and GPUs with lower precisions highlights ADON’s efficiency and potential in robust predictive modeling for large-scale Earth system computations.

Bai, Zhe↗
DOE OSTI2024

An Ethics-Based Review of Generative Artificial Intelligence: Assuring Responsible Use (Version 1.0)

The rapid expansion of generative artificial intelligence (GenAI) has generated excitement regarding its potential benefits and concern over its ethical implications. Governments, corporations, and standards organizations have described ethical principles to direct GenAI's development and use; however, practical guidance for implementing these principles is limited. Addressing this gap is critical, especially considering the array of risks associated with GenAI, such as legal liabilities, privacy concerns, security threats, and potential misuse. Robust policies and procedures are critical to support responsible deployment of GenAI. This report examines Pacific Northwest National Laboratory (PNNL)’s approach to promoting responsible GenAI use. Proposed initiatives include developing policies based on ethical principles, creating a governance process to review projects relative to those principles, and implementing onboarding processes for training staff. The governance framework described in this report adapts the structure and principles of Institutional Review Boards (IRBs), traditionally used in human subjects research, for GenAI ethical review, providing oversight. Ethical principles guiding responsible GenAI usage include transparency and accountability, privacy, fairness, safety, security, and validity and reliability. To operationalize these principles, we propose forming a GenAI Assurance Council (GAC) that mirrors the IRB's structure. The GAC will evaluate GenAI projects across privacy, accountability, transparency, safety, security, fairness, and validity dimensions. Complementing policy and governance is AI literacy training to support staff understanding of GenAI's ethical implications. An initial training effort for AI Incubator Chat—a GenAI tool deployed at PNNL—showed promising results, underscoring the importance of clear guidelines and user accountability. Collaborative efforts and the dissemination of best practices are also discussed. The proposed GAC model and AI literacy training provide a blueprint for establishing ethical GenAI use and governance, offering practical tools to bridge the gap between ethical principles and real-world applications. The responsible integration of GenAI at PNNL entails a multifaceted approach involving policy development, ethical governance, and AI literacy training. The positive initial feedback and collaborative opportunities position PNNL to lead by example in GenAI's responsible use, reflecting a proactive stance in addressing the ethical, legal, and societal challenges associated with this emerging technology. PNNL's systematic and ethical approach to GenAI offers a model for other institutions to emulate, promoting safe and responsible technological advancements in the AI domain.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Facet-dependent structure and dissociation of water at pristine IrO 2 /water interfaces

Understanding the microscopic structure of water at metal oxide interfaces is crucial for advancing electrocatalysis. IrO 2 , specifically, has shown exceptional activity for electrochemical water oxidation, but we currently lack a fundamental understanding of how the surface structure of IrO 2 impacts water reactivity. In this work, we developed a machine learning potential trained to first-principles accuracy for modeling IrO 2 /water interfaces across different facets: (110), (100), (101), and (001). Using extensive machine learning molecular dynamics simulations, we investigated the spontaneous dissociation of water molecules at these interfaces. Our results reveal a distinct dissociation probability trend: (110) > (100) ≈ (101) > (001), which we attribute primarily to the reaction thermodynamics of surface water dissociation. A strong correlation is observed between the surface Ir–O bond distances and the dissociation probabilities, highlighting the role of surface geometry in modulating reactivity. As a consequence, the interfacial solvation structures and hydrogen bonding environments are dynamically tuned by the varying water dissociation capabilities across facets. This work elucidates how water dissociation energetics depend on surface orientation and interfacial structure, offering atomistic insights into manipulating reaction chemistry at electrocatalytic interfaces.

organic↗
DOE OSTI2023

Development of a machine-learning-based ionic-force correction model for quantum molecular dynamic simulations of warm dense matter

In this work Δ learning is used to map orbital-free density functional theory (OF-DFT) ionic forces to the corresponding Kohn-Sham (KS) DFT ionic forces. The development of the approximate force difference in terms of the ion positions is constructed and serves as a stand in for the ground truth force difference. Descriptor vectors for ion configurations are constructed using all distance between ions in conjunction with an indexing based on a nearest neighbor ranking. It is demonstrated that such a scheme of descriptors can uniquely describe an ionic configuration up to a rotation and reflection when no ambiguity in the nearest neighbor ranking exists. How to handle the case when an ambiguity exists in the nearest neighbor ranking is discussed. As a proof of principle, the model is trained and tested on warm dense hydrogen at temperatures between 1 and 15 eV. Once tested, the model was used to perform molecular dynamic simulations of warm dense hydrogen. Furthermore, the resulting energies and pressures are within 1% and 2% of their respective target KS values.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Automating the Analysis of Large Language Models Responses through Zero-Shot Question Answering

Recent advancements in Large Language Models (LLMs) have shown significant potential in various applications, yet their evaluation, particularly in zero-shot question answering scenarios, remains a challenging task. In this study, our objective was to explore precision metrics for Large Language Models (LLM) and design and implement a software pipeline to automatically evaluate LLMs' outputs under zero-shot question answering. Zero-shot question answering involves a model providing answers to questions about topics it hasn't seen during training. It leverages the principles of zero-shot learning by relying on semantic understanding and generalization from related knowledge. The data used was metadata from medical databases on congenital heart disease. We explored eleven LLM metrics and selected three for our evaluation: BLEU, BERTScore, and MoverScore. BLEU calculates a score based on the overlap of n-grams (contiguous sequences of n items, typically words) between the machine-generated translation and the reference translations. Higher BLEU scores indicate better correspondence between the machine-generated and human-generated translations. BERTScore is a metric used to evaluate the quality of machine-generated text by measuring the similarity of token embeddings produced by BERT (Bidirectional Encoder Representations from Transformers) between the generated text and reference text. MoverScore is a metric that quantifies the dissimilarity between the distributions of word embeddings from machine-generated text and reference text, emphasizing semantic similarity over exact token overlap. We also introduced HBKI, a composite metric summarizing these approaches. We tested five models —GPT-3, Llama-2, Gemini 1.5 Pro, Solar 10.7B, and Mixtral-8x7b. Our software pipeline, designed and implemented using Object-Oriented Programming principles, allows users to customize the selection and extraction of features for topics of interest in their own research. Our results show that MoverScore delivered the most precise evaluation of the LLM's outputs, while Mixtral-8x7b achieved the best overall performance in extracting metadata from the databases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2015

Building international genomics collaboration for global health security

Genome science and technologies are transforming life sciences globally in many ways and becoming a highly desirable area for international collaboration to strengthen global health. The Genome Science Program at the Los Alamos National Laboratory is leveraging a long history of expertise in genomics research to assist multiple partner nations in advancing their genomics and bioinformatics capabilities. The capability development objectives focus on providing a molecular genomics-based scientific approach for pathogen detection, characterization, and biosurveillance applications. The general approaches include introduction of basic principles in genomics technologies, training on laboratory methodologies and bioinformatic analysis of resulting data, procurement, and installation of next-generation sequencing instruments, establishing bioinformatics software capabilities, and exploring collaborative applications of the genomics capabilities in public health. Genome centers have been established with public health and research institutions in the Republic of Georgia, Kingdom of Jordan, Uganda, and Gabon; broader collaborations in genomics applications have also been developed with research institutions in many other countries.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Machine learning molecular dynamics simulations toward exploration of high-temperature properties of nuclear fuel materials: case study of thorium dioxide

Predicting materials properties of nuclear fuel compounds is a challenging task in materials science. Their thermodynamical behaviors around and above the operational temperature are essential for the design of nuclear reactors. However, they are not easy to measure, because the target temperature range is too high to perform various standard experiments safely and accurately. Moreover, theoretical methods such as first-principles calculations also suffer from the computational limitations in calculating thermodynamical properties due to their high calculation-costs and complicated electronic structures stemming from f-orbital occupations of valence electrons in actinide elements. Here, we demonstrate, for the first time, machine-learning molecular-dynamics to theoretically explore high-temperature thermodynamical properties of a nuclear fuel material, thorium dioxide. The target compound satisfies first-principles calculation accuracy because f-electron occupation coincidentally diminishes and the scheme meets sampling sufficiency because it works at the computational cost of classical molecular-dynamics levels. We prepare a set of training data using first-principles molecular dynamics with small number of atoms, which cannot directly evaluate thermodynamical properties but captures essential atomistic dynamics at the high temperature range. Then, we construct a machine-learning molecular-dynamics potential and carry out large-scale molecular-dynamics calculations. Consequently, we successfully access two kinds of thermodynamic phase transitions, namely the melting and the anomalous λ transition induced by large diffusions of oxygen atoms. Furthermore, we quantitatively reproduce various experimental data in the best agreement manner by selecting a density functional scheme known as SCAN. Our results suggest that the present scale-up simulation-scheme using machine-learning techniques opens up a new pathway on theoretical studies of not only nuclear fuel compounds, but also a variety of similar materials that contain both heavy and light elements, like thorium dioxide.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2023

Dynamics of Aqueous Electrolyte Solutions: Challenges for Simulations

This Perspective article focuses on recent simulation work on the dynamics of aqueous electrolytes. It is well-established that full-charge, nonpolarizable models for water and ions generally predict solution dynamics that are too slow in comparison to experiments. Models with reduced (scaled) charges do better for solution diffusivities and viscosities but encounter issues describing other dynamic phenomena such as nucleation rates of crystals from solution. Polarizable models show promise, especially when appropriately parametrized, but may still miss important physical effects such as charge transfer. First-principles calculations are starting to emerge for these properties that are in principle able to capture polarization, charge transfer, and chemical transformations in solution. Finally, while direct ab initio simulations are still too slow for simulations of large systems over long time scales, machine-learning models trained on appropriate first-principles data show significant promise for accurate and transferable modeling of electrolyte solution dynamics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Fundamental microscopic properties as predictors of large-scale quantities of interest: Validation through grain boundary energy trends

Correlations between fundamental microscopic properties computable from first principles, which we term canonical properties, and complex large-scale quantities of interest (QoIs) provide an avenue to predictive materials discovery. Here, we propose that such correlations can be efficiently discovered through simulations utilizing approximate interatomic potentials (IPs), which serve as an ensemble of “synthetic materials”. As a proof of principle we build a regression model relating canonical properties to the symmetric tilt grain boundary (GB) energy curves in face-centered cubic crystals, characterized by the scaling factor in the universal lattice matching model of Runnels et al. (2016), which we take to be our QoI. Our analysis recovers known correlations of GB energy to other properties and discovers new ones. We also demonstrate, using available density functional theory (DFT) GB energy data, that the regression model constructed from IP data is consistent with DFT results, confirming the assumption that the IPs and DFT belong to same statistical pool and thereby validating the approach. Regression models constructed in this fashion can be used to predict large-scale QoIs based on first-principles data and provide a general method for training IPs for QoIs beyond the scope of first-principles calculations.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Safe Operations at Roadway Junctions: Intelligent Roadway Infrastructure as Functional Interlocking

Automated vehicle (AV) technology is quickly maturing, and the corresponding infrastructure systems that evaluate traffic and communicate to vehicles requires sophisticated sensing and perception technologies, referred to as intelligent roadway infrastructure (IRI), to complement emerging AV capabilities. IRI provides signals to vehicles, indicating right-of-way for vehicles and communicating to approaching AVs that no other vehicle is failing to yield. This capability, denoted as safety-affirmative signaling, provides a green light or a green arrow as appropriate and affirms through communication links to connected vehicles when it is safe to proceed. About 36% of collisions occur at intersections, with most occurring upon left turns (22.2%) or crossing over (12.6%), and only a small percentage (1.2%) while turning right at an intersection. Of all intersection crashes about half (52.5%) of those vehicles were traveling through a signalized intersection 2. Safety-affirmative signaling would guarantee safety of AV fleet vehicles, by providing the interlocking principle, a term from automated train control that only allows progression through a railway intersection after affirming no opportunity for a crash exists. IRI through safety-affirmative signaling would bring performance and safety to complex roadway intersections where AV transit fleet service is most needed, as well as safety benefits to traditional, non-automated vehicles and vulnerable road users. The implementation of IRI has functional, programmatic, and technical challenges. Research work performed at the National Renewable Energy Laboratory (NREL) in an integrative approach encapsulating these themes, and termed infrastructure perception and control (IPC) is motivated by improved performance (travel time), improved safety (reduced collisions), and improved energy efficiency (less fuel burned and minimized production of greenhouse gases). IPC is intended not only for roadway and intersection applications but also in extension to inform complementary buildings and grid systems to enable better co-management, as vehicles and their charging needs become increasingly integrated into the built environment. The NREL IPC project presents an open-source framework, architecture, and supporting technology to implement IRI, addressing critical issues such as fusion of data, reliability, standardization of data interfaces, and confidence of detection. The framework is informed by previous experience in U.S. Department of Defense research technology, specifically in the use of radar to detect, identify, and track aerial threats. These principles combined with multi-sensor fusion provides for a complete digital twin with known and measurable confidence and accuracy from which safety-affirmative signaling can be developed and deployed.

ADVANCED PROPULSION SYSTEMS,MATHEMATICS AND COMPUT↗
DOE OSTI2016

Idaho National Laboratory Technology to Market (T2M) Final Report

The objective of this project is to establish a tiered Technology to Market (T2M) curriculum for basic researchers to project leads to measure the effect of technology transfer skills on project success and impact. The plan will train five researchers in basic technology transfer principles where success will be measured by assessing improvements in T2M skills and knowledge after the training is complete, likely using before and after surveys.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2021

Safe Operations at Roadway Junctions - Design Principles from Automated Guideway Transit

Herein this paper describes a system-level view of a fully automated transit system comprising a fleet of automated vehicles (AVs) in driverless operation, each with an SAE level 4 Automated Driving System, along with its related safety infrastructure and other system equipment. This AV system-level control is compared to the automatic train control system used in automated guideway transit technology, particularly that of communications-based train control (CBTC). Drawing from the safety principles, analysis methods, and risk assessments of CBTC systems, comparable functional subsystem definitions are proposed for AV fleets in driverless operation. With the prospect of multiple AV fleets operating within a single automated mobility district, the criticality of protecting roadway junctions requires an approach like that of automated fixed-guideway transit systems, in which a guideway switch zone "interlocking" at each junction location deconflicts railway traffic, affirming safe passage. The analogous AV protection safety subsystem is defined as fail-safe equipment that monitors roadway intersections and junctions, communicates traffic signal status, perceives and communicates alerts and signals to AV connected vehicles concerning potential unsafe conditions, and performs related primary safety functions. Conclusions are drawn that the AV protection roadway intersection functions must be performed by local roadside equipment dedicated to protecting each roadway intersection and junction. Further, it is concluded that the communications technology connecting the infrastructure with the vehicle to perform this vital, fail-safe protection should meet specific functional and performance criteria.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2023

Uncertainty quantification of a deep learning fuel property prediction model

Deep learning models are being widely used in the field of combustion. Given the black-box nature of typical neural network based models, uncertainty quantification (UQ) is critical to ensure the reliability of predictions as well as the training datasets, and for a principled quantification of noise and its various sources. Deep learning surrogate models for predicting properties of chemical compounds and mixtures have been recently shown to be promising for enabling data-driven fuel design and optimization, with the ultimate goal of improving efficiency and lowering emissions from combustion engines. In this study, UQ is performed for a multi-task deep learning model that simultaneously predicts the research octane number (RON), Motor Octane Number (MON), and Yield Sooting Index (YSI) of pure components and multicomponent blends. The deep learning model is comprised of three smaller networks: Extractor 1, Extractor 2, and Predictor, and a mixing operator. The molecular fingerprints of individual components are encoded via Extractor 1 and Extractor 2, the mixing operator generates fingerprints for mixtures/blends based on linear mixing operation, and the predictor maps the fingerprint to the target properties. Two different classes of UQ methods, Monte Carlo ensemble methods and Bayesian neural networks (BNNs), are employed for quantifying the epistemic uncertainty. Combinations of Bernoulli and Gaussian distributions with DropConnect and DropOut techniques are explored as ensemble methods. All the DropConnect, DropOut and Bayesian layers are applied to the predictor network. Aleatoric uncertainty is modeled by assuming that each data point has an independent uncertainty associated with it. The results of the UQ study are further analyzed to compare the performance of BNN and ensemble methods. Although this study is confined to UQ of fuel property prediction, the methodologies are applicable to other deep learning frameworks that are being widely used in the combustion community.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2021

White Box Access to Quantum Testbeds for Co-Design

At Lawrence Livermore National Laboratory (LLNL), we operate and maintain the Quantum Device and Integration Testbed (QuDIT) facility, a small quantum testbed that supports about 10 active research teams (including our own) and over 50 internal and external collaborators. This testbed is designed to give remote white box access to users for research, training, and outreach. A guiding principle behind the development of our testbed infrastructure, software and user interfaces is to empower users to perform experiments at the cutting edge of quantum information science at any level of abstraction, from materials studies, device physics and control and characterization techniques to algorithm development and quantum operating system design. Our testbed targets a multilevel quantum system (qudit) to expand the accessible Hilbert space of a simple-to-manufacture quantum device and focuses on quantum simulation, typically implemented through custom gates designed with quantum optimal control methods, rather than on a universal computing framework with a fixed gate set. We leverage the Lab’s high-performance computing (HPC) program and related expertise to simulate quantum systems, develop hybrid algorithms, and generate gates optimized for given simulations. Additionally, we have adopted a co-design philosophy from the HPC community in designing new hardware, so that the systems we develop are optimized for the specific physics simulations we plan to use them for.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗