Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “training”

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 1,603 records · Page 89
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

TensorFlow at Scale: Performance and productivity analysis of distributed training with Horovod, MLSL, and Cray PE ML

Summary Deep learning has proven to be a successful tool for solving a large variety of problems in various scientific fields and beyond. In recent years, the models as well as the available datasets have grown bigger and more complicated, and thus, an increasing amount of computing resources is required in order to train these models in a reasonable amount of time. Besides being able to use HPC resources, deep learning model developers want flexible frameworks which allow for rapid prototyping. One of the most important of these frameworks is Google TensorFlow, which provides both features, ie, good performance as well as flexibility. In this paper, we discuss different solutions for scaling the TensorFlow Framework to thousands of nodes on contemporary Cray XC supercomputing systems.

Kurth, Thorsten↗
DOE OSTI2023

Changes in Distal Tibial Microarchitecture During Eight Weeks of U.S. Army Basic Combat Training Differ by Sex and Race

ABSTRACT Basic combat training (BCT) is a physically rigorous period at the beginning of a soldier's career that induces bone formation in the tibia. Race and sex are determinants of bone properties in young adults but their influences on changes in bone microarchitecture during BCT are unknown. The purpose of this work was to determine the influence of sex and race on changes in bone microarchitecture during BCT. Bone microarchitecture was assessed at the distal tibia via high‐resolution peripheral quantitative computed tomography at the beginning and end of 8 weeks of BCT in a multiracial cohort of trainees (552 female, 1053 male; mean ± standard deviation [SD] age = 20.7 ± 3.7 years) of which 25.4% self‐identified as black, 19.5% as race other than black or white (other races combined), and 55.1% as white. We used linear regression models to determine whether changes in bone microarchitecture due to BCT differed by race or sex, after adjusting for age, height, weight, physical activity, and tobacco use. We found that trabecular bone density (Tb.BMD), thickness (Tb.Th), and volume (Tb.BV/TV), as well as cortical BMD (Ct.BMD) and thickness (Ct.Th) increased following BCT in both sexes and across racial groups (+0.32% to +1.87%, all p < 0.01). Compared to males, females had greater increases in Tb.BMD (+1.87% versus +1.40%; p = 0.01) and Tb.Th (+0.87% versus +0.58%; p = 0.02), but smaller increases in Ct.BMD (+0.35% versus +0.61%; p < 0.01). Compared to black trainees, white trainees had greater increases in Tb.Th (+0.82% versus +0.61%; p = 0.03). Other races combined and white trainees had greater increases in Ct.BMD than black trainees (+0.56% and + 0.55% versus +0.32%; both p ≤ 0.01). Changes in distal tibial microarchitecture, consistent with adaptive bone formation, occur in trainees of all races and sexes, with modest differences by sex and race. Published 2023. This article is a U.S. Government work and is in the public domain in the USA. JBMR Plus published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research.

Hughes, Julie M.↗
DOE OSTI2019

Creating Training Data for Scientific Named Entity Recognition with Minimal Human Effort

Scientific Named Entity Referent Extraction is often more complicated than traditional Named Entity Recognition (NER). For example, in polymer science, chemical structure may be encoded in a variety of nonstandard naming conventions, and authors may refer to polymers with conventional names, commonly used names, labels (in lieu of longer names), synonyms, and acronyms. As a result, accurate scientific NER methods are often based on task-specific rules, which are difficult to develop and maintain, and are not easily generalized to other tasks and fields. Machine learning models require substantial expert-annotated data for training. Here we propose polyNER: a semi-automated system for efficient identification of scientific entities in text. PolyNER applies word embedding models to generate entity-rich corpora for productive expert labeling, and then uses the resulting labeled data to bootstrap a context-based word vector classifier. Evaluation on materials science publications shows that the polyNER approach enables improved precision or recall relative to a state-of-the-art chemical entity extraction system at a dramatically lower cost: it required just two hours of expert time, rather than extensive and expensive rule engineering, to achieve that result. This result highlights the potential for human-computer partnership for constructing domain-specific scientific NER systems.

Crowdsourcing↗
DOE OSTI2016

A program to generate simulated radioxenon beta–gamma data for concentration verification and validation and training exercises

Abstract Several hundred simulated radioxenon beta-gamma data files were developed to assist in evaluating the performance and results from radioxenon concentration calculation analysis at the International Data Center (IDC) and other National Data Centers (NDC). PNNL developed a Beta-Gamma Simulator (BGSim) that incorporated GEANT-modeled data sets from radioxenon decay chains, as well as functionality to use nuclear detector-acquired data sets to create new beta-gamma spectra with varying amounts of background, 133Xe, 131mXe, 133mXe, 135Xe, and 222Rn and its decay products. The program has been implemented on a web-based applications platform and allows the user to create very specific data sets that incorporate most of the operational parameters for the current beta-gamma systems deployed in the International Monitoring System (IMS) and the On-site Inspection (OSI) equipment. After an initial beta-gamma simulations program was developed, additional uses began to be identified for the program output: training sets of two-dimensional spectra for data analysts at the IDC and other NDC, spectra for exercises such as the Integrated Field Exercise 2014 (IFE14) held in Jordan at the Dead Sea, and testing new analysis methods and algorithms

custom datasets↗
DOE OSTI2023

Biomarkers of oxidative stress, diet and exercise distinguish soldiers selected and non-selected for special forces training

The metabolomic profiles of Soldiers entering the U.S. Special Forces Assessment and Selection course (SFAS) have not been evaluated. To compare pre-SFAS blood metabolomes of Soldiers selected during SFAS versus those not selected, and explore the relationships between the metabolome, physical performance, and diet quality. Fasted blood samples and food frequency questionnaires were collected from 761 Soldiers prior to entering SFAS to assess metabolomic profiles and diet quality, respectively. Physical performance was assessed throughout SFAS. Between-group differences (False Discovery Rate < 0.05) in 108 metabolites were detected. Selected candidates had higher levels of compounds within xenobiotic, pentose phosphate, and corticosteroid metabolic pathways, while non-selected candidates had higher levels of compounds potentially indicative of oxidative stress (i.e., sphingomyelins, acylcarnitines, glutathione, amino acids). Multiple compounds higher in non-selected versus selected candidates included: 1-carboxyethylphenylalanine; 4-hydroxy-nonenal-glutathione; α-hydroxyisocaproate; hexanoylcarnitine; sphingomyelin and were associated with lower diet quality and worse physical performance. Candidates selected during SFAS had higher pre-SFAS levels of circulating metabolites that were associated with resistance to oxidative stress, higher physical performance and higher diet quality. In contrast, non-selected candidates had higher levels of metabolites potentially indicating elevated oxidative stress. These findings indicate that Soldiers who were selected for continued Special Forces training enter the SFAS course with metabolites associated with healthier diets and better physical performance. Additionally, the non-selected candidates had higher levels of metabolites that may indicate elevated oxidative stress, which could result from poor nutrition, non-functional overreaching/overtraining, or incomplete recovery from previous physical activity.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

An Open Combinatorial Diffraction Dataset Including Consensus Human and Machine Learning Labels with Quantified Uncertainty for Training New Machine Learning Models

Modern machine learning and autonomous experimentation schemes in materials science rely on accurate analysis of the data ingested by these models. Unfortunately, accurate analysis of the underlying data can be difficult, even for domain experts, complicating the training of the models intended to drive experiments. This is especially true when the goal is to identify the presence of weak signatures in diffraction or spectroscopic datasets. In this work, we examine a set of as-obtained diffraction data that track the phase transition from monoclinic to tetragonal in a Nb-doped VO2 film as a function of temperature and dopant concentration. We then task a set of domain experts and a set of machine learning experts with identifying which phase is present in each diffraction pattern manually and algorithmically, respectively; in both cases, the labels can vary dramatically, especially at the phase boundaries. We use the mode of the labels and the Shannon entropy as a method to capture, preserve and propagate consensus labels and their variance. Further we use the expert labels as a benchmark and demonstrate the use of Shannon entropy weighted scoring to test the performance of machine learning generated labels. Finally, we propose a material data challenge centered around generating improved labeling algorithms. This real-world dataset curated with expert labels can act as test bed for new algorithms. The raw data, annotations and code used in this study are all available online at data.gov and the interested reader is encouraged to replicate and improve the existing models

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Respecting causality for training physics-informed neural networks

While the popularity of physics-informed neural networks (PINNs) is steadily rising, to this date PINNs have not been successful in simulating dynamical systems whose solution exhibits multi-scale, chaotic or turbulent behavior. In this work we attribute this shortcoming to the inability of existing PINNs formulations to respect the spatio-temporal causal structure that is inherent to the evolution of physical systems. We argue that this is a fundamental limitation and a key source of error that can ultimately steer PINN models to converge towards erroneous solutions. We address this pathology by proposing a simple re-formulation of PINNs loss functions that can explicitly account for physical causality during model training. We demonstrate that this simple modification alone is enough to introduce significant accuracy improvements, as well as a practical quantitative mechanism for assessing the convergence of a PINNs model. Here we provide state-of-the-art numerical results across a series of benchmarks for which existing PINNs formulations fail, including the chaotic Lorenz system, the Kuramoto–Sivashinsky equation in the chaotic regime, and the Navier–Stokes equations. To the best of our knowledge, this is the first time that PINNs have been successful in simulating such systems, introducing new opportunities for their applicability to problems of industrial complexity.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

A systematic method for selecting molecular descriptors as features when training models for predicting physiochemical properties

Machine learning has proven to be a powerful tool for accelerating biofuel development. Although numerous models are available to predict a range of properties using chemical descriptors, there is a trade-off between interpretability and performance. Neural networks provide predictive models with high accuracy at the expense of some interpretability, while simpler models such as linear regression often lack in accuracy. In addition to model architecture, feature selection is also critical for developing interpretable and accurate predictive models. We present a method for systematically selecting molecular descriptor features and developing interpretable machine learning models without sacrificing accuracy. Our method simplifies the process of selecting features by reducing feature multicollinearity and enables discoveries of new relationships between global properties and molecular descriptors. To demonstrate our approach, we developed models for predicting melting point, boiling point, flash point, yield sooting index, and net heat of combustion with the help of the Tree-based Pipeline Optimization Tool (TPOT). For training, we used publicly available experimental data for up to 8351 molecules. Our models accurately predict various molecular properties for organic molecules (mean absolute percent error (MAPE) ranges from 3.3% to 10.5%) and provide a set of features that are well-correlated to the property. This method enables researchers to explore sets of features that significantly contribute to the prediction of the property, offering new scientific insights. To help accelerate early stage biofuel research and development, we also integrated the data and models into a open-source, interactive web tool.

09 BIOMASS FUELS↗
DOE OSTI2016

Solving the master equation without kinetic Monte Carlo: Tensor train approximations for a CO oxidation model

In multiscale modeling of heterogeneous catalytic processes, one crucial point is the solution of a Markovian master equation describing the stochastic reaction kinetics. Usually, this is too high-dimensional to be solved with standard numerical techniques and one has to rely on sampling approaches based on the kinetic Monte Carlo method. In this study we break the curse of dimensionality for the direct solution of the Markovian master equation by exploiting the Tensor Train Format for this purpose. The performance of the approach is demonstrated on a first principles based, reduced model for the CO oxidation on the RuO{sub 2}(110) surface. We investigate the complexity for increasing system size and for various reaction conditions. The advantage over the stochastic simulation approach is illustrated by a problem with increased stiffness.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2018

Gradient-based optimization for regression in the functional tensor-train format

Predictive analysis of complex computational models, such as uncertainty quantification (UQ), must often rely on using an existing database of simulation runs. In this paper we consider the task of performing low-multilinear-rank regression on such a database. Specifically we develop and analyze an efficient gradient computation that enables gradient-based optimization procedures, including stochastic gradient descent and quasi-Newton methods, for learning the parameters of a functional tensor-train (FT). We compare our algorithms with 22 other nonparametric and parametric regression methods on 10 real-world data sets and show that for many physical systems, exploiting low-rank structure facilitates efficient construction of surrogate models. Here, we use a number of synthetic functions to build insight into behavior of our algorithms, including the rank adaptation and group-sparsity regularization procedures that we developed to reduce overfitting. Finally we conclude the paper by building a surrogate of a physical model of a propulsion plant on a naval vessel.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2018

Radiation Therapy Patient Education Review and a Case Study Using the Virtual Environment for Radiotherapy Training System

The emergence of modern learning environments for radiation therapy (RT) education offers innovative opportunities for RT patients. This article presents a descriptive review of the current state of practice for two recently available tools, the “Virtual Environment for Radiotherapy Training” (VERT) and the “Patient Education And Radiotherapy Learning” (PEARL) systems. Subsequently, a case study summarizes the instructional design and development of an RT breast cancer patient education program, with the intention of providing a blueprint for further patient education initiatives that incorporate VERT.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2018

Verification and validation of atmospheric transport models with the UF Training Reactor

The University of Florida Training Reactor (UFTR) is a 100-kW ARGONAUT type reactor with a unique design such that a large volume of air passes through the core, producing and releasing significant quantities of 41 Ar. Once released, the radioargon is dispersed throughout the atmosphere and decays by a 1294 keV gamma ray emission. This quantified, validated, and controlled 41 Ar plume can be utilized to verify and validate atmospheric transport models (ATMs). This is accomplished by coupling an ATM with MCNP and comparing the results to experimental measurements. For this study, an in-house developed Gaussian dispersion model and the EPA approved Gaussian model AERMOD were coupled with MCNP to simulate the results of a six hour measurement taken on May 8th, 2018. For the measurements and simulations, a grid of seven NaI detectors were placed in four different rooftop locations near the UFTR plume stack. With easterly winds dominating during the morning and early afternoon, both the measurement results and simulations displayed a higher 41 Ar photon count rate on the detectors west of the plume stack throughout the day. Overall, the AERMOD simulations produced better results when compared to the in-house model, particularly for the detectors placed east and west of the plume stack. While some of the AERMOD data showed a strong correlation with the measurements, the results remain inconsistent. Here, due to the varying accuracy of the simulation results, it is theorized that the relationship between measurement results and simulations will improve with the use of localized weather information or the utilization of non-Gaussian dispersion models.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2023

Characterization of transverse electron pulse trains using RF powered traveling wave metallic comb striplines

Advancements in ultrafast electron microscopy have allowed elucidation of spatially selective structural dynamics. However, as the spatial resolution and imaging capabilities have made progress, quantitative characterization of the electron pulse trains has not been reported at the same rate. In fact, inexperienced users have difficulty replicating the technique because only a few dedicated microscopes have been characterized thoroughly. Systems replacing laser driven photoexcitation with electrically driven deflectors especially suffer from a lack of quantified characterization because of the limited quantity. The primary advantages to electrically driven systems are broader frequency ranges, ease of use and simple synchronization to electrical pumping. Here, we characterize the technical parameters for electrically driven UEM including the shape, size and duration of the electron pulses using low and high frequency chopping methods. At high frequencies, pulses are generated by sweeping the electron beam across a chopping aperture. For low frequencies, the beam is continuously forced off the optic axis by a DC potential, then momentarily aligned by a countering pulse. Using both methods, we present examples that measure probe durations of 2 ns and 10 ps for the low and high frequency techniques, respectively. Here, we also discuss how the implementation of a pulsed probe affects STEM imaging conditions by adjusting the first condenser lens.

36 MATERIALS SCIENCE↗
DOE OSTI2019

Beryllium-10 dating of the Foothills Erratics Train in Alberta, Canada, indicates detachment of the Laurentide Ice Sheet from the Rocky Mountains at ~15 ka

The Foothills Erratics Train consists of large quartzite blocks of Rocky Mountains origin deposited on the eastern slopes of the Rocky Mountain Foothills in Alberta between ~53.5°N and 49°N. The blocks were deposited in their present locations when the western margin of the Laurentide Ice Sheet (LIS) detached from the local ice masses of the Rocky Mountains, which initiated the opening of the southern end of the ice-free corridor between the Cordilleran Ice Sheet and the LIS. In this study, we use 10 Be exposure dating to constrain the beginning of this decoupling. Based on a group of 12 samples well-clustered in time, we date the detachment of the western LIS margin from the Rocky Mountain front to ~14.9 ± 0.9 ka. This is ~1000 years later than previously assumed, but a lack of a latitudinal trend in the ages over a distance of ~500 km is consistent with the rapid opening of a long wedge of unglaciated terrain portrayed in existing ice-retreat reconstructions. Finally, a later separation of the western LIS margin from the mountain front implies higher ice margin–retreat rates in order to meet the Younger Dryas ice margin position near the boundary of the Canadian Shield ~2000 years later.

10Be exposure dating↗
DOE OSTI2019

ff19SB: Amino-acid specific protein backbone parameters trained against quantum mechanics energy surfaces in solution

Molecular dynamics (MD) simulations have become increasingly popular in studying the motions and functions of biomolecules. The accuracy of the simulation, however, is highly determined by the molecular mechanics (MM) force field (FF), a set of functions with adjustable parameters to compute the potential energies from atomic positions. However, the overall quality of the FF, such as our previously published ff99SB and ff14SB, can be limited by assumptions that were made years ago. In the updated model presented here (ff19SB), we have significantly improved the backbone profiles for all 20 amino acids. We fit coupled Φ/ψ parameters using 2D Φ/ψ conformational scans for multiple amino acids, using as reference data the entire 2D quantum mechanics (QM) energy surface. We address the polarization inconsistency during dihedral parameter fitting by using both QM and MM in solution. Finally, we examine possible dependency of the backbone fitting on side chain rotamer. To extensively validate ff19SB parameters, we have performed a total of ~5 milliseconds MD simulations in explicit solvent. Here, our results show that after amino-acid specific training against QM data with solvent polarization, ff19SB not only reproduces the differences in amino acid specific Protein Data Bank (PDB) Ramachandran maps better, but also shows significantly improved capability to differentiate amino acid dependent properties such as helical propensities. We also conclude that an inherent underestimation of helicity is present in ff14SB, which is (inexactly) compensated by an increase in helical content driven by the TIP3P bias toward overly compact structures. In summary, ff19SB, when combined with a more accurate water model such as OPC, should have better predictive power for modeling sequence-specific behavior, protein mutations, and also rational protein design.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2018

Computational Seismology Workshop Trains Early-Career Scientists

The Computational Infrastructure for Geodynamics (CIG), in collaboration with Lawrence Livermore National Laboratory (LLNL), held a workshop last year that focused on computational seismology. The workshop provided training in seismic waveform processing, visualization, and high-performance computing (HPC) waveform simulation. Fifty-five predominantly early-career participants—graduate students and postdocs—from the United States and 16 other countries attended. The workshop was the first of its kind to feature full access to HPC resources for research-grade example problems.

58 GEOSCIENCES↗
DOE OSTI2021

An Observationally Trained Markov Model for MJO Propagation

A Markovian stochastic model is developed for studying the propagation of the Madden-Julian Oscillation (MJO). This model represents the daily changes in real time multivariate MJO (RMM) indices as random functions of their current state and background conditions. The probability distribution function of the RMM changes is obtained using a machine learning algorithm trained to maximize MJO forecast skills using observed daily indices of RMM and different modes of variability. Skillful forecasts are obtained for lead times between 8 and 27 days. Large ensemble simulations by the stochastic model show that with monsoonal changes in the background state, MJO propagation across the Maritime Continent (MC) is most likely to be disrupted in boreal spring and summer when MJO events propagate from favorable conditions over the Indian Ocean to unfavorable ones over the MC, and predictability is higher during spring and summer when MJO activity is away from the MC region.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Monitoring Seismic Velocity Changes Across the San Jacinto Fault Using Train‐Generated Seismic Tremors

Abstract Microseismic noise has been used for seismic velocity monitoring. However, such signals are dominated by low‐frequency surface waves that are not ideal for detecting changes associated with small tectonic processes. Here we show that it is possible to extract stable, high‐frequency body waves using seismic tremors generated by freight trains. Such body waves allow us to focus on small velocity perturbations in the crust with high spatial resolution. We report on 10 years of seismic velocity temporal changes at the San Jacinto Fault. We observe and map a two‐month‐long episode of velocity changes with complex spatial distribution and interpret the velocity perturbation as produced by a previously undocumented slow‐slip event. We verify the hypothesis through numerical simulations and locate this event along a fault segment believed to be locked. Such a slow‐slip event stresses its surroundings and may trigger a major earthquake on a fault section approaching failure.

58 GEOSCIENCES↗