Search NASA⌕ Search

SEARCH · Search NASA

Results for “embedding model”

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 OSTI2019

Nonequilibrium Environment Dynamics in a Frequency-Dependent Polarizable Embedding Model

Hybrid quantum mechanical/molecular mechanical (QM/MM) models are some of the most powerful and computationally feasible approaches to account for solvent effects or more general environmental perturbations on quantum chemical systems. In their more recent formulations (known as polarizable embedding) they can account for electrostatic and mutual polarization effects between the QM and the MM subsystems. In this paper, a polarizable embedding scheme based on induced dipoles that is both able to describe electron evolution of the embedded QM system in an efficient manner as well as to capture the frequency dependent behavior of the solvent, is proposed. The effects of this frequency dependent solvent on a time- dependent model system - the Rabi oscillations of H+2 in a resonant - field are considered. The solvent is shown to introduce only mild perturbations when the excitation frequencies of the solvent and the solute are off-resonant. However, the dynamics of the H+2 are fundamentally changed in the presence of a near-resonant excitation solvent.

Wildman, Andrew↗
DOE OSTI2021

A Bayesian Calibration Framework with Embedded Model Error for Model Diagnostics

We study the utility and performance of a Bayesian model error embedding construction in the context of molecular dynamics modeling of metallic alloys, where we embed model error terms in existing interatomic potential model parameters. To alleviate the computational burden of this approach, we propose a framework combining likelihood approximation and Gaussian process surrogates. Here we leverage sparse Gaussian process techniques to construct a hierarchy of increasingly accurate but more expensive surrogate models. This hierarchy is then exploited by multilevel Markov chain Monte Carlo methods to efficiently sample from the target posterior distribution. We illustrate the utility of this approach by calibrating an interatomic potential model for a family of gold-copper alloys. In particular, this case study highlights effective means for dealing with computational challenges with Bayesian model error embedding in large-scale physical models, and the utility of embedded model error for model diagnostics.

Bayesian inference↗
DOE OSTI2024

PhysBERT: A text embedding model for physics scientific literature

The specialized language and complex concepts in physics pose significant challenges for information extraction through Natural Language Processing (NLP). Central to effective NLP applications is the text embedding model, which converts text into dense vector representations for efficient information retrieval and semantic analysis. In this work, we introduce PhysBERT, the first physics-specific text embedding model. Pre-trained on a curated corpus of 1.2 × 106 arXiv physics papers and fine-tuned with supervised data, PhysBERT outperforms leading general-purpose models on physics-specific tasks, including the effectiveness in fine-tuning for specific physics subdomains.

Hellert, Thorsten (ORCID:0000000227970926)↗
DOE OSTI2025

Domain-specific text embedding model for accelerator physics

Accelerator physics presents unique challenges for natural language processing (NLP) due to its specialized terminology and complex concepts. A key component in overcoming these challenges is the development of robust text embedding models that transform textual data into dense vector representations, facilitating efficient information retrieval and semantic understanding. In this work, we introduce AccPhysBERT, a sentence embedding model fine-tuned specifically for accelerator physics. Our model demonstrates superior performance across a range of downstream NLP tasks, surpassing existing models in capturing the domain-specific nuances of the field. We further showcase its practical applications, including semantic paper-reviewer matching and integration into retrieval-augmented generation systems, highlighting its potential to enhance information retrieval and knowledge discovery in accelerator physics. Published by the American Physical Society 2025

Hellert, Thorsten (ORCID:0000000227970926)↗
DOE OSTI2025

Data and Code for Understanding Generative AI Content with Embedding Models

This repository contains code for the experiments in the paper "Understanding Generative AI Content with Embedding Models". Constructing high-quality features is critical to any quantitative data analysis. While feature engineering was historically addressed by carefully hand-crafting data representations based on domain expertise, deep neural networks (DNNs) now offer a radically different approach. DNNs implicitly engineer features by transforming their input data into hidden feature vectors called embeddings. For embedding vectors produced by foundation models -- which are trained to be useful across many contexts -- we demonstrate that simple and well-studied dimensionality-reduction techniques such as Principal Component Analysis uncover inherent heterogeneity in input data concordant with human-understandable explanations. Of the many applications for this framework, we find empirical evidence that there is intrinsic separability between real samples and those generated by artificial intelligence (AI).

Vargas, Max [Pacific Northwest National Laboratory↗
DOE OSTI2021

Harmonic Infrared and Raman Spectra in Molecular Environments Using the Polarizable Embedding Model

We present a fully analytic approach to calculate infrared (IR) and Raman spectra of molecules embedded in complex molecular environments modeled using the fragment-based polarizable embedding (PE) model. We provide the theory for the calculation of analytic second-order geometric derivatives of molecular energies and first-order geometric derivatives of electric dipole moments and dipole–dipole polarizabilities within the PE model. The derivatives are implemented using a general open-ended response theory framework, thus allowing for an extension to higher-order derivatives. The embedding-potential parameters used to describe the environment in the PE model are derived through first-principles calculations, thus allowing a wide variety of systems to be modeled, including solvents, proteins, and other large and complex molecular environments. Here, we present proof-of-principle calculations of IR and Raman spectra of acetone in different solvents. This work is an important step toward calculating accurate vibrational spectra of molecules embedded in realistic environments.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2013

Seismic Soil-Structure Interaction Analyses of a Deeply Embedded Model Reactor – SASSI Analyses

This report summarizes the SASSI analyses of a deeply embedded reactor model performed by BNL and CJC and Associates, as part of the seismic soil-structure interaction (SSI) simulation capability project for the NEAMS (Nuclear Energy Advanced Modeling and Simulation) Program of the Department of Energy. The SASSI analyses included three cases: 0.2 g, 0.5 g, and 0.9g, all of which refer to nominal peak accelerations at the top of the bedrock. The analyses utilized the modified subtraction method (MSM) for performing the seismic SSI evaluations. Each case consisted of two analyses: input motion in one horizontal direction (X) and input motion in the vertical direction (Z), both of which utilized the same in-column input motion. Besides providing SASSI results for use in comparison with the time domain SSI results obtained using the DIABLO computer code, this study also leads to the recognition that the frequency-domain method should be modernized so that it can better serve its mission-critical role for analysis and design of nuclear power plants.

42 ENGINEERING↗
DOE OSTI2019

Embedded Model Error Representation for Bayesian Model Calibration

Model error estimation remains one of the key challenges in uncertainty quantification and predictive science. For computational models of complex physical systems, model error, also known as structural error or model inadequacy, is often the largest contributor to the overall predictive uncertainty. This work builds on a recently developed framework of embedded, internal model correction, in order to represent and quantify structural errors, together with model parameters,within a Bayesian inference context. We focus specifically on a Polynomial Chaos representation with additive modification of existing model parameters, enabling a non-intrusive procedure for efficient approximate likelihood construction, model error estimation, and disambiguation of model and data errors’ contributions to predictive uncertainty. The framework is demonstrated on several synthetic examples, as well as on a chemical ignition problem.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2018

Embedded Model Error Representation for Bayesian Model Calibration

Model error estimation remains one of the key challenges in uncertainty quantification and predictive science. For computational models of complex physical systems, model error, also known as structural error or model inadequacy,is often the largest contributor to the overall predictive uncertainty. This work builds on a recently developed framework of embedded, internal model correction, in order to represent and quantify structural errors, together with model parameters, within a Bayesian inference context. We focus specifically on a Polynomial Chaos representation with additive modification of existing model parameters, enabling a non-intrusive procedure for efficient approximate likelihood construction, model error estimation, and disambiguation of model and data errors' contributions to predictive uncertainty. As a result, the framework is demonstrated on several synthetic examples, as well as on a chemical ignition problem

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

A crystal plasticity finite element model embedding strain-rate sensitivities inherent to deformation mechanisms: Application to alloy AZ31

We report the fundamental power-law relationship representing the flow rule in crystal visco-plasticity ensures uniqueness in the selection of slip systems accommodating imposed plastic strain-rates. The power-law relationship also introduces an artificially high strain-rate sensitivity in crystal plasticity simulations, unless a high value of the power-law exponent is used. However, the use of high values for the exponent is limited by numerical tractability. This paper presents a numerical method implemented in a crystal plasticity finite element (CPFE) model for embedding any value of the power-law exponent reflecting the true material strain-rate sensitivity. Importantly, the method does not increase computation time involved in the simulations. The enhanced CPFE model is used to interpret and predict a complex strain-rate sensitive response and microstructural evolution of AZ31 Mg alloy. Measured values of strain-rate sensitivity for slip and twinning modes are used in the simulations. Calculations show that the model successfully captures the phenomena pertaining to the effect of changing applied strain-rate on the mechanical response including flow stress and evolution of texture and twinning for a broad range of strain-rates ranging from 10 -3 s -1 to 10 3 s -1 and loading orientations in tension and compression. It is shown that such predictions are a consequence of not only relative amounts of slip and twinning activities driven by a set of accurately characterized hardening law parameters but also values of the strain-rate sensitivities inherent to individual deformation mechanisms. Besides, the model validates the measured strain-rate dependency of deformation mechanisms while accurately reproducing the mechanical data. Hence, the model can be used to verify and further refine or infer measured strain-rate sensitivity per deformation mechanism by reproducing experimental data.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Formulation of a Consistent Multi‐Species Canopy Description for Hydrodynamic Models Embedded in Large‐Scale Land‐Surface Representations of Mixed‐Forests

Abstract The plant hydrodynamic approach represents a recent advancement to land surface modeling, in which stomatal conductance responds to water availability in the xylem rather than in the soil. To provide a realistic representation of tree hydrodynamics, hydrodynamic models must resolve processes at the level of a single modeled tree, and then scale the resulting fluxes to the canopy and land surface. While this tree‐to‐canopy scaling is trivial in a homogeneous canopy, mixed‐species canopies require careful representation of the species properties and a scaling approach that results in a realistic description of both the canopy and individual‐tree hydrodynamics, as well as leaf‐level fluxes from the canopy and their forcing. Here, we outline advantages and pitfalls of three commonly used approaches for representing mixed forests in land surface models and present a new framework for scaling vegetation characteristics and fluxes in mixed forests. The new formulation scales fluxes from the tree to canopy level in an energy‐ and mass‐conservative way and allows for a consistent multi‐species/multi‐type canopy description by hydrodynamic models.

Bohrer, G.↗
DOE OSTI2024

Understanding Generative AI Content with Embedding Models

The construction of high-quality numerical features is critical to any quantitative data analysis. Feature engineering has been historically addressed by carefully hand-crafting data representations based on domain expertise. This work views the internal representations of modern deep neural networks (DNNs), called embeddings, as an implicit form of traditional feature engineering. For trained DNNs, we show that these embeddings can reveal interpretable, high-level concepts in unstructured sample data. We use these embeddings in natural language and computer vision tasks to uncover both inherent heterogeneity in the underlying data and human-understandable explanations for it. In particular, we find empirical evidence that there is inherent separability between real data and those generated from AI models.

Vargas, Max↗