Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “classification models”

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 1,027 records · Page 57
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

AI Model Benchmarking for Nonproliferation Applications: Steel Thread Benchmarking Task Force Technical Report (Rev. 2)

Steel Thread is a NA-22 venture that seeks to build trustworthy, reliable AI models that can be used in a wide variety of nonproliferation tasks. A key aspect of building these models is developing appropriate benchmarks and evaluation methods, which will enable the venture to identify and adapt models to provide the most value in the nonproliferation domain. Benchmarks must be relevant to key tasks in this domain, such as question answering, information retrieval, document summarization and classification, consensus analysis, and image and data analysis. This report 1) provides an overview of benchmark design, evaluation, and challenges; 2) reviews a variety of open benchmarks, with a focus on language models and tasks; and 3) identifies benchmarks that are most relevant to Steel Thread. This report is intended to serve as a basis for further efforts to classify and evaluate benchmarks and their correlation with success on nonproliferation-specific tasks. The Steel Thread venture has defined benchmarks to be a particular combination of a dataset (or datasets) and a metric (or metrics) conceptualized as representing one or more specific tasks or sets of abilities for a specific modality. It is adopted by a research community as a shared framework for comparing methods.1 It includes 1) Data: Labeled (a designated subset not used for training, which could be all the data), 2) Metric: A way to quantify performance, 3) Task/Ability: What the benchmark is testing, 4) Protocol: A structured and repeatable evaluation process, 5) Baseline/Reference Model: For comparison; could be statistical, rule-based, SME-derived, or another model, and 6) Maintenance Plan: to update with new information over time; important for long-term utility. For further clarity, the definition includes what a benchmark, in this context, is not. It is not a corpus of training data, specific to a model (it is intended to apply to a range of models), a universal evaluation of performance, a guarantee that the ‘top’ model on the leaderboard will be the best fit for every specific use case, an all-encompassing proof of a model’s universal quality, nor is it a one-size-fits-all measure of success. It does not cover every real-world constraint (like operational, ethical, or cost considerations), a systems integration test, or a unit test. This definition was inspired by and resulted from discussions within the Steel Thread Benchmarking Task Force. This group was formed to define what we would mean as a benchmark within Steel Thread but persisted as the need to develop a thorough understanding of the large and expanding existing benchmarking space. This technical report is a result of the group’s divide and conquer approach to exploring this space. The release of benchmarks might not be progressing as quickly as model development, but it is moving very fast, as many benchmarks quickly become saturated, when state-of-the-art models score so close to the benchmark’s ceiling that their results are virtually indistinguishable. At that point, the test no longer differentiates between new systems, so researchers usually stop reporting scores as the benchmark no longer informs about improvements from the next generation of models. In the OpenAI announcement of GPT-5, they reported results on six flagship public benchmarks (AIME 2025, SWE-bench Verified, Aider Polyglot, MMMU, HealthBench Hard, GPQA) but the full system-card covers roughly thirty-five separate evaluations, comprising hundreds of test task items in total. There have been some efforts to summarize benchmarks in specific fields, like for text-to-image generation, but these surveys have had a narrow methodology scope. Therefore, a comprehensive survey of all benchmarks or even all benchmarks that could be relevant to Steel Thread is outside of the scope of this report. We chose some specific benchmarks to investigate in detail.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Computational catalyst discovery: Active classification through myopic multiscale sampling

We report the recent boom in computational chemistry has enabled several projects aimed at discovering useful materials or catalysts. We acknowledge and address two recurring issues in the field of computational catalyst discovery. First, calculating macro-scale catalyst properties is not straightforward when using ensembles of atomic-scale calculations [e.g., density functional theory (DFT)]. We attempt to address this issue by creating a multi-scale model that estimates bulk catalyst activity using adsorption energy predictions from both DFT and machine learning models. The second issue is that many catalyst discovery efforts seek to optimize catalyst properties, but optimization is an inherently exploitative objective that is in tension with the explorative nature of early-stage discovery projects. In other words, why invest so much time finding a “best” catalyst when it is likely to fail for some other, unforeseen problem? We address this issue by relaxing the catalyst discovery goal into a classification problem: “What is the set of catalysts that is worth testing experimentally?” Here, we present a catalyst discovery method called myopic multiscale sampling, which combines multiscale modeling with automated selection of DFT calculations. It is an active classification strategy that seeks to classify catalysts as “worth investigating” or “not worth investigating” experimentally. Our results show an ~7–16 times speedup in catalyst classification relative to random sampling. These results were based on offline simulations of our algorithm on two different datasets: a larger, synthesized dataset and a smaller, real dataset.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Vulnerability of Wetlands Due to Projected Sea-Level Rise in the Coastal Plains of the South and Southeast United States

Coastal wetlands are vulnerable to accelerated sea-level rise, yet knowledge about their extent and distribution is often limited. We developed a land cover classification of wetlands in the coastal plains of the southern United States along the Gulf of Mexico (Texas, Louisiana, Mississippi, Alabama, and Florida) using 6161 very-high (2 m per pixel) resolution WorldView-2 and WorldView-3 satellite images from 2012 to 2015. Area extent estimations were obtained for the following vegetated classes: marsh, scrub, grass, forested upland, and forested wetland, located in elevation brackets between 0 and 10 m above sea level at 0.1 m intervals. Sea-level trends were estimated for each coastal state using tide gauge data collected over the period 1983–2021 and projected for 2100 using the trend estimated over that period. These trends were considered conservative, as sea level rise in the region accelerated between 2010 and 2021. Estimated losses in vegetation area due to sea level rise by 2100 are projected to be at least 12,587 km 2 , of which 3224 km 2 would be coastal wetlands. Louisiana is expected to suffer the largest losses in vegetation (80%) and coastal wetlands (75%) by 2100. Such high-resolution coastal mapping products help to guide adaptation plans in the region, including planning for wetland conservation and coastal development.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2012

Building a Shared Definitional Model of Long Duration Human Spaceflight

Objective: To establish the need for a shared definitional model of long duration human spaceflight, that would provide a framework and vision to facilitate communication, research and practice In 1956, on the eve of human space travel, Hubertus Strughold first proposed a "simple classification of the present and future stages of manned flight" that identified key factors, risks and developmental stages for the evolutionary journey ahead. As we look to new destinations, we need a current shared working definitional model of long duration human space flight to help guide our path. Here we describe our preliminary findings and outline potential approaches for the future development of a definition and broader classification system

Arias, Diana↗
DOE OSTI2024

A Deep Multimodal Representation Learning Framework for Accurate Molecular Properties Prediction

Drug discovery is a complex and challenging process, requiring the optimization of candidate compounds to identify those with the potential to become safe and effective drugs. Predicting molecular properties is an indispensable step in the drug discovery pipeline. Traditionally, this process is costly and time-intensive, involving multiple rounds of experiments and clinical trials, rendering it impractical for every candidate compound. Deep learning techniques have emerged as a promising approach to drug discovery to reduce the cost and time required to identify novel drugs. However, prevalent research in deep learning models focused on predicting molecular properties has primarily fixated on single-modal models, which utilize a single modality of data, neglecting the potential benefits of combining different data modalities. To overcome this limitation, we introduce MRL-Mol: a deep \textbf{M}ultimodal \textbf{R}epresentation \textbf{L}earning framework for accurate \textbf{Mol}ecular properties prediction. MRL-Mol harnesses three data modalities: sequence, graph, and image, augmenting the depth of comprehension. Leveraging a large-scale unlabeled dataset~($\sim$1M unique molecules), we pretrain MRL-Mol to extract inter- and intra-modal information. Our study demonstrates the superior performance of MRL-Mol in predicting molecular properties across six benchmark datasets, including both classification and regression tasks. Notably, MRL-Mol outperforms other state-of-the-art molecular properties prediction models. These findings suggest that by combining information from multiple data modalities, MRL-Mol can comprehend molecules better than single-modal deep learning models and identify molecular properties with better accuracy.

Yang, Yuxin↗
DOE OSTI2020

Right-handed neutrino dark matter, neutrino masses, and non-standard cosmology in a 2HDM

Here we explore the dark matter phenomenology of a weak-scale right-handed neutrino in the context of a Two Higgs Doublet Model. The expected signal at direct detection experiments is different from the usual spin-independent and spin-dependent classification since the scattering with quarks depends on the dark matter spin. The dark matter relic density is set by thermal freeze-out and in the presence of non-standard cosmology, where an Abelian gauge symmetry is key for the dark matter production mechanism. We show that such symmetry allows us to simultaneously address neutrino masses and the flavor problem present in general Two Higgs Doublet Model constructions. Lastly, we outline the region of parameter space that obeys collider, perturbative unitarity and direct detection constraints.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2021

Classification and computation of extreme events in turbulent combustion

In the design of practical combustion systems, ensuring safety and reliability is an important requirement. For instance, reliably avoiding lean blowout, flame flashback or inlet unstart is critical for ensuring safe operation. Currently, the science of predicting such events is based on prior experience, limited modeling or diagnostic tools and purely statistical approaches. Even though computational and experimental tools for studying combustion devices have vastly advanced in the last three decades, the analysis of such failure events has not been pursued widely. While the use of data for model development and calibration is being widely accepted, the extension to failure events introduces numerous challenges. In particular, the focus here is on so-called data-poor problems, where the cost of generating data is extremely high and is not easily amenable to existing computational and experimental approaches. Data-poor problems are particularly relevant when related to extreme events (also called anomalous events) that can lead to catastrophic failure of the system. It is argued that transient events that describe such failure can have different causal mechanisms. To develop the scientific inference process, a classification of such problems is used to determine specific modeling paths as well as computational tools needed. Research opportunities in the emerging field of extreme event prediction are highlighted in order to identify critical and immediate needs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Is tokenization needed for masked particle modeling?

In this work, we significantly enhance masked particle modeling (MPM), a self-supervised learning scheme for constructing highly expressive representations of unordered sets relevant to developing foundation models for high-energy physics. In MPM, a model is trained to recover the missing elements of a set, a learning objective that requires no labels and can be applied directly to experimental data. We achieve significant performance improvements over previous work on MPM by addressing inefficiencies in the implementation and incorporating a more powerful decoder. We compare several pre-training tasks and introduce new reconstruction methods that utilize conditional generative models without data tokenization or discretization. We show that these new methods outperform the tokenized learning objective from the original MPM on a new test bed for foundation models for jets, which includes using a wide variety of downstream tasks relevant to jet physics, such as classification, secondary vertex finding, and track identification.

conditional generative models↗
NASA NTRS1993

Tree Classification Software

This paper introduces the IND Tree Package to prospective users. IND does supervised learning using classification trees. This learning task is a basic tool used in the development of diagnosis, monitoring and expert systems. The IND Tree Package was developed as part of a NASA project to semi-automate the development of data analysis and modelling algorithms using artificial intelligence techniques. The IND Tree Package integrates features from CART and C4 with newer Bayesian and minimum encoding methods for growing classification trees and graphs. The IND Tree Package also provides an experimental control suite on top. The newer features give improved probability estimates often required in diagnostic and screening tasks. The package comes with a manual, Unix 'man' entries, and a guide to tree methods and research. The IND Tree Package is implemented in C under Unix and was beta-tested at university and commercial research laboratories in the United States.

Buntine, Wray↗
NASA NTRS2017

Sensitivity Analysis and Uncertainty Quantification of a Mars Ascent Vehicle Concept

The design of a conceptual Mars ascent vehicle is a challenging problem. In order to aid the vehicle and mission concept design it is important to understand the driving design parameters and the expected performance in the presence of model errors and uncertainties. An existing six degree of freedom simulation model is analyzed on a statistical basis using the methods available in the Design Analysis Kit for Optimization and Terascale Applications toolkit. The methods utilized include conventional Monte Carlo techniques, metamodeling via polynomial chaos expansions, and global variance-based sensitivity analyses. Two additional analysis methods referred to as “Monte Carlo filtering” and ”Classification trees” are used to determine which uncertain parameters are driving the performance of the vehicle. Monte Carlo filtering provides a methodology to determine which parameters cause qualitatively different behavior while the classification trees use heuristics to partition the input space and assign probabilities to each partition. These methods serve as qualitative descriptors of model sensitivity while variance-based global sensitivity analysis seeks a quantitative mapping from total output variance to the variance of individual inputs. Application of these techniques to several outputs of a Mars ascent vehicle concept simulation indicates that only a select few input factors dominate their variance.

Noyes, Connor↗
DOE OSTI2024

Simulating water dynamics related to pedogenesis across space and time: Implications for four-dimensional digital soil mapping

Digital soil mapping (DSM) relies on machine-learning and geostatistics to represent soil property observations across space. DSM techniques are powerful but often empirical, being limited to the quality and density of point samples. Water dynamics are closely related to soil variability, and the physics that govern water movement are well known. Hydrological properties can hence be simulated by physical models through space and time, unveiling key characteristics about soils. We propose the use of hydrologic models to map soils across the surface (2D), depth (1D), and time (1D)–which provides a 4D approach to digital soil mapping (4DSM). The Distributed Hydrology Soil Vegetation Model (DHSVM) was applied to a watershed currently under pasture. Moisture sensors and wells were installed at different depths in the watershed on summit, sideslope and toeslope positions to validate the model. DHSVM simulations of soil moisture distribution and depth to saturation were performed during the hydrological year (October 2008-September 2009). Clusters of similar pixels based on soil moisture values were determined using Dynamic Time Warping (DTW) to align temporal data and K-means. Clustering was performed both seasonally and for the entire year. Temporal patterns simulated by DHSVM matched measurements given by moisture sensors and wells. Seasonal clusters differed from the annual cluster. Distinct clusters were observed for each season and with depth, showing that spatiotemporal soil variability is lost when statically assessing soils. Spatiotemporal clusters corroborated field observations of fragipan occurrence not explicitly spatially mapped by Soil Survey Geographic Database (SSURGO). If a connection can be made between water and soils, static and dynamic soil variability can be predicted using physically based hydrologic models. Hydrologic models can benefit soil mapping by enabling reliable 4D simulation of water dynamics, which are fundamental to soil variability and soil classification and directly relate to biological, physical and chemical soil processes not captured by typical soil sampling protocols.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2019

Evaluation of Matthews-WLR-CH4: A New Wetland, Lake, and Reservoir Methane Emissions Data Set

Methane (CH4) is emitted from a variety of sources, both natural and anthropogenic, and is the second most important greenhouse gas contributing to climate change. Natural wetlands are the largest single contributor to annual global CH4 emissions and other inland water sources, such as lakes and reservoirs, produce CH4 but have received less attention. These emission sources are among the most uncertain components of the global CH4 cycle. These uncertainties stem from numerous issues, including the vast variability in these ecosystems and the sensitivity of CH4 emissions to interannual climate variations. Existing CH4 emission models display large differences in wetland spatial extent and include simplified, or lack all together, wetland-type classifications. Furthermore, CH4 emissions from lakes and reservoirs are typically combined with wetlands, or ignored, on a regional and global scale. Our NASA Interdisciplinary Research in Earth Science (IDS) study is focused on developing a suite of data sets on wetlands, lakes, and reservoirs (WLR) including global distributions of types, CH4-centric classifications, and daily CH4 emission rates (hereafter Matthews-WLR-CH4). This project is unique in that it comprises source-independent emission data for WLR, the first spatially-explicit data set of lake and reservoir CH4 emissions, and wetland emissions comprising numerous different wetland-type classifications.The focus of this study is to perform the initial evaluation of the wetland and lake CH4 emissions data from Matthews-WLR-CH4. This evaluation will focus on the domain of North America and include: 1) estimates of the contribution of wetland and lake sources to total CH4 emissions and atmospheric mixing ratios, 2) comparisons of wetland fraction and CH4 emissions to a suite of existing models (e.g., Wetland and Wetland CH4 Inter-comparison of Models Project (WETCHIMP), Global Carbon Project (GCP)), and 3) an accuracy assessment with in situ airborne and tower-based atmospheric CH4 observations when the Matthews-WLR-CH4 emissions are implemented in forward and inverse CarbonTracker-Lagrange (CT-L) CH4 model simulations. Here we will present the new Matthews-WLR-CH4 wetland and lake CH4 emissions data, the initial evaluation of these products, and the path forward for our project.

Reservoir↗
DOE OSTI2021

DeepMerge – II. Building robust deep learning algorithms for merging galaxy identification across domains

In astronomy, neural networks are often trained on simulation data with the prospect of being used on telescope observations. Unfortunately, training a model on simulation data and then applying it to instrument data leads to a substantial and potentially even detrimental decrease in model accuracy on the new target dataset. Simulated and instrument data represent different data domains, and for an algorithm to work in both, domain-invariant learning is necessary. Here we employ domain adaptation techniques— Maximum Mean Discrepancy (MMD) as an additional transfer loss and Domain Adversarial Neural Networks (DANNs)— and demonstrate their viability to extract domain-invariant features within the astronomical context of classifying merging and non-merging galaxies. Additionally, we explore the use of Fisher loss and entropy minimization to enforce better in-domain class discriminability. We show that the addition of each domain adaptation technique improves the performance of a classifier when compared to conventional deep learning algorithms. We demonstrate this on two examples: between two Illustris-1 simulated datasets of distant merging galaxies, and between Illustris-1 simulated data of nearby merging galaxies and observed data from the Sloan Digital Sky Survey. The use of domain adaptation techniques in our experiments leads to an increase of target domain classification accuracy of up to ~20%. With further development, these techniques will allow astronomers to successfully implement neural network models trained on simulation data to efficiently detect and study astrophysical objects in current and future large-scale astronomical surveys.

galaxies: interactions↗
DOE OSTI2022

Detecting Quantum Critical Points of Correlated Systems by Quantum Convolutional Neural Network Using Data from Variational Quantum Eigensolver

Machine learning has been applied to a wide variety of models, from classical statistical mechanics to quantum strongly correlated systems, for classifying phase transitions. The recently proposed quantum convolutional neural network (QCNN) provides a new framework for using quantum circuits instead of classical neural networks as the backbone of classification methods. We present the results from training the QCNN by the wavefunctions of the variational quantum eigensolver for the one-dimensional transverse field Ising model (TFIM). We demonstrate that the QCNN identifies wavefunctions corresponding to the paramagnetic and ferromagnetic phases of the TFIM with reasonable accuracy. The QCNN can be trained to predict the corresponding ‘phase’ of wavefunctions around the putative quantum critical point even though it is trained by wavefunctions far away. The paper provides a basis for exploiting the QCNN to identify the quantum critical point.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
NASA NTRS2011

Design and Impacts of Land-Biogenic-Atmosphere Coupling in the NASA-Unified WRF (NU-WRF) Modeling System

Land-Atmosphere coupling is typically designed and implemented independently for physical (e.g. water and energy) and chemical (e.g. biogenic emissions and surface depositions)-based models and applications. Differences in scale, data requirements, and physics thus limit the ability of Earth System models to be fully coupled in a consistent manner. In order for the physical-chemical-biological coupling to be complete, treatment of the land in terms of surface classification, condition, fluxes, and emissions must be considered simultaneously and coherently across all components. In this study, we investigate a coupling strategy for the NASA-Unified Weather Research and Forecasting (NU-WRF) model that incorporates the traditionally disparate fluxes of water and energy through NASA's LIS (Land Information System) and biogenic emissions through BEIS (Biogenic Emissions Inventory System) and MEGAN (Model of Emissions of Gases and Aerosols from Nature) into the atmosphere. In doing so, inconsistencies across model inputs and parameter data are resolved such that the emissions from a particular plant species are consistent with the heat and moisture fluxes calculated for that land cover type. In turn, the response of the atmospheric turbulence and mixing in the planetary boundary layer (PBL) acts on the identical surface type, fluxes, and emissions for each. In addition, the coupling of dust emission within the NU-WRF system is performed in order to ensure consistency and to maximize the benefit of high-resolution land representation in LIS. The impacts of those self-consistent components on' the simulation of atmospheric aerosols are then evaluated through the WRF-Chem-GOCART (Goddard Chemistry Aerosol Radiation and Transport) model. Overall, this ambitious project highlights the current difficulties and future potential of fully coupled. components. in Earth System models, and underscores the importance of the iLEAPS community in supporting improved knowledge of processes and innovative approaches for models and observations.

Tan, Qian↗
DOE OSTI2024

Physics-inspired spatiotemporal-graph AI ensemble for the detection of higher order wave mode signals of spinning binary black hole mergers

We present a new class of AI models for the detection of quasi-circular, spinning, non-precessing binary black hole mergers whose waveforms include the higher order gravitational wave modes ($\ell$, |m|) = {(2,2), (2,1), (3,3), (3,2), (4,4)}, and mode mixing effects in the $\ell$ = 3, |m| = 2 harmonics. These AI models combine hybrid dilated convolution neural networks to accurately model both short- and long-range temporal sequential information of gravitational waves; and graph neural networks to capture spatial correlations among gravitational wave observatories to consistently describe and identify the presence of a signal in a three detector network encompassing the Advanced LIGO and Virgo detectors. We first trained these spatiotemporal-graph AI models using synthetic noise, using 1.2 million modeled waveforms to densely sample this signal manifold, within 1.7 h using 256 NVIDIA A100 GPUs in the Polaris supercomputer at the Argonne Leadership Computing Facility. This distributed training approach exhibited optimal classification performance, and strong scaling up to 512 NVIDIA A100 GPUs. With these AI ensembles we processed data from a three detector network, and found that an ensemble of 4 AI models achieves state-of-the-art performance for signal detection, and reports two misclassifications for every decade of searched data. We distributed AI inference over 128 GPUs in the Polaris supercomputer and 128 nodes in the Theta supercomputer, and completed the processing of a decade of gravitational wave data from a three detector network within 3.5 h. Finally, we fine-tuned these AI ensembles to process the entire month of February 2020, which is part of the O3b LIGO/Virgo observation run, and found 6 gravitational waves, concurrently identified in Advanced LIGO and Advanced Virgo data, and zero false positives. This analysis was completed in one hour using one NVIDIA A100 GPU.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Predictive Indicators of the Performance of Large Language Models

In several mission contexts, it is desirable to estimate the performance of large language models (LLMs) on tasks that we cannot run directly. In light of published “scaling laws” our hypothesis is that some tasks should be consistently more challenging than others based on characteristics of the task. The goal of this project was to begin quantifying how much information about LLM performance can be gained from the features of a model and a task. Two of our statistical models struggled to converge. Pass/fail test results may provide limited information for inference beyond model quality and task difficulty, but we see no evidence at this time for significant feature interaction effect sizes, arguing for simple models. Future work extending the models to capitalize on perplexity of ground truth answers is suggested. This project also introduces “Depth of Knowledge Variant Testing” as a strategy for more finely assessing language models on open domain question and answer tasks. We developed sets of questions that ask a language model to produce similar information while demonstrating increasing depth of knowledge, and also relabeled existing Q&A test questions with their depth of knowledge. Our results suggest further consideration of Bloom’s taxonomy and further refinement of prompts to properly elicit information at varying depths. In the course of this work, we set up a basic infrastructure for standardizing tasks and testing many language models on these tasks. In addition to testing the predictive quality of model features and performance across test suites, with this project we have introduced two new task features to contextualize each test question: the Dewey Classification main category of information covered, and the Bloom’s taxonomy level that corresponds to the depth of knowledge probed by the question. Splits across these and other features produced over five hundred task subtypes with distinct feature vectors, which we tested on half a dozen models.

97 MATHEMATICS AND COMPUTING↗