Search NASA⌕ Search

SEARCH · Search NASA

Results for “data representations”

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 1,711 records · Page 95
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

Machine Learning Explainability and Transferability for Path Navigation

Deep neural networks are powerful tools for machine perception. Unfortunately their decisions are difficult to explain due to the complexity and size of the networks. Previously we have alleviated this issue by using the representational portion of a deep neural network and combining it with a k-nearest neighbor (KNN) classifier. Through inspection of the decisions made by the KNN, we can directly see the training data responsible for the decisions, allowing us to determine the quality of the overall decision and the quality of the representational layer of the deep NN. While the technique worked well, it requires tens of thousands of latent vectors to be stored for classification. In addition, it lacks the ability to show how parts of an image influence the classification decision. Here we address these issues by 1) Using a radial basis function network (RBFN) in place of the KNN allowing far fewer images to be used in deployment and 2) Using an auto encoder network for explainability. In addition to these techniques, we examine the effects of transfer learning to determine that results are robust. All results are tested on a domain where an unmanned aerial vehicle (UAV) navigates a forest trail through a single camera.

explainable machine learning↗
DOE OSTI2022

Advanced Hydropower and PSH Capacity Expansion Modeling (Final Report on HydroWIRES D1 Improvements to Capacity Expansion Modeling)

Hydropower and pumped-storage hydropower (PSH) have played a key role in providing flexible, low-carbon electricity to the U.S. electricity system for over a century. As variable generation (VG) deployment increases the demand for flexible, dispatchable generation, it is important to use all available methods for understanding how hydropower and PSH interfaces with VG in a future low-carbon grid. Capacity expansion models (CEMs) of electricity systems are often used to study future electricity scenarios, but these tools often have difficulty representing site- and technology-details of hydropower and PSH due to limited spatial, temporal, or process resolution. This report demonstrates a set of model advancements to improve hydropower and PSH representations in CEMs and other models that consider hydropower and PSH's role in electricity systems. Model advancements include new data integration to define closed-loop PSH resource availability and cost, along with site-level parameterizations of existing PSH capacity and energy storage specifications. Shifting energy across seasons for both hydropower and PSH is explored to demonstrate the potential value of long-duration storage. New representations of hydropower upgrades offer opportunities to increase dispatchability, add pumps, or independently add capacity or energy depending on what is most valuable. New model structures and data are demonstrated individually and in combination to observe their impact on model results and provide initial insights into what is most important for analysts, electricity system planners, and hydropower decision-makers to consider when assessing future roles of hydropower. Improving flexibility of existing hydropower assets has the potential to reduce CO 2 emissions through complementing VG technologies and/or improve electricity system economics by reducing the need to deploy higher-cost systems. Large-scale energy storage is also shown to provide opportunities for balancing seasonal differences in VG, particularly solar photovoltaics (PV). Initial analysis with a new closed-loop PSH resource and cost dataset also demonstrate the potential for new PSH deployment to offer new grid flexibility opportunities. Future analysis and modeling of hydropower's role in the U.S. and other electricity systems could include methodological improvements based on those discussed here along with sensitivity analysis to better represent current and future potential hydropower and PSH flexibility. Improvements to geospatial and site-level data can further improve the understanding of how hydropower and PSH can be upgraded and operated in response to future electricity system needs. This work lays the groundwork to use advanced planning models to explore the range of roles hydropower and PSH can play in the grid of the future.

13 HYDRO ENERGY↗
DOE OSTI2026

Leveraging large language models to address data scarcity in machine learning for graphene synthesis

Machine learning in experimental materials science faces significant challenges due to the scarcity of data, which are costly and time-consuming to generate, particularly when relying on in-house experiments. Literature data mining offers a potential solution but introduces issues like mixed data quality, inconsistent formats, and non-uniform reporting of synthesis parameters, resulting in partially missing and heterogeneous features across the dataset. Here, we propose data imputation and feature engineering methods that employ pre-trained large language models (LLMs) to enhance machine learning performance on scarce, heterogeneous datasets, demonstrated on graphene CVD synthesis data and the ML-HydPARK hydrogen storage dataset. GPT models perform data imputation via tailored prompting and semantic normalization of inconsistently reported features through embeddings, for example, to harmonize the complex nomenclature of CVD substrates. Beyond yielding more diverse and richer feature representations than traditional methods such as K-nearest neighbors (KNN) and Multivariate Imputation by Chained Equations (MICE), LLM-based data imputation is evaluated against dataset characteristics and prompting strategies. We vary the level of autonomy granted to the LLM, from generic prompting that leverages pre-trained knowledge for autonomous data generation to data-informed prompting that constrains outputs using target-specific information, and demonstrate which level of autonomy yields superior imputation performance across datasets and feature types. The proposed data engineering methods markedly improve downstream performance; for example, in graphene layer number classification using a support vector machine (SVM), binary accuracy increases from 39% to 65% and ternary accuracy from 52% to 72%. Fine-tuning experiments on both datasets show that combining our proposed LLM-based data imputation and feature encoding methods with numerical machine learning predictors outperforms standalone fine-tuned LLM predictors in data-scarce settings. The proposed strategies emphasize data enhancement techniques rather than refining learning architectures or regularizing loss functions, offering a broadly applicable framework for improving machine learning performance on scarce, inhomogeneous datasets.

Chemical vapor deposition↗
NASA NTRS1993

Process and representation in graphical displays

Our initial model of graphic comprehension has focused on statistical graphs. Like other models of human-computer interaction, models of graphical comprehension can be used by human-computer interface designers and developers to create interfaces that present information in an efficient and usable manner. Our investigation of graph comprehension addresses two primary questions: how do people represent the information contained in a data graph?; and how do they process information from the graph? The topics of focus for graphic representation concern the features into which people decompose a graph and the representations of the graph in memory. The issue of processing can be further analyzed as two questions: what overall processing strategies do people use?; and what are the specific processing skills required?

Gillan, Douglas J.↗
DOE OSTI2021

Data-driven enhancement of coherent structure-based models for predicting instantaneous wall turbulence

Predictions of the spatial representation of instantaneous wall-bounded flows, via coherent structure-based models, are highly sensitive to the geometry of the representative structures employed by them. In this study, we propose a methodology to extract the three-dimensional (3-D) geometry of the statistically significant eddies from multi-point wall-turbulence datasets, for direct implementation into these models to improve their predictions. The methodology is employed here for reconstructing a 3-D statistical picture of the inertial wall coherent turbulence for all canonical wall-bounded flows, across a decade of friction Reynolds number (Re T ). These structures are responsible for the Re T -dependence of the skin-friction drag and also facilitate the inner-outer interactions, making them key targets of structure-based models. The empirical analysis brings out the geometric self-similarity of the large-scale wall-coherent motions and also suggests the hairpin packet as the representative flow structure for all wall-bounded flows, thereby aligning with the framework on which the attached eddy model (AEM) is based. The same framework is extended here to also model the very-large-scaled motions, with a consideration of their differences in internal versus external flows. Implementation of the empirically-obtained geometric scalings for these large structures into the AEM is shown to enhance the instantaneous flow predictions for all three velocity components. Finally, an active flow control system driven by the same geometric scalings is conceptualized, towards favourably altering the influence of the wall coherent motions on the skin-friction drag.

42 ENGINEERING↗
DOE OSTI2022

TULIP: An RNA-seq-based Primary Tumor Type Prediction Tool Using Convolutional Neural Networks

Background: With cancer as one of the leading causes of death worldwide, accurate primary tumor type prediction is critical in identifying genetic factors that can inhibit or slow tumor progression. There have been efforts to categorize primary tumor types with gene expression data using machine learning, and more recently with deep learning, in the last several years. Methods In this paper, we developed four 1-dimensional (1D) Convolutional Neural Network (CNN) models to classify RNA-seq count data as one of 17 highly represented primary tumor types or 32 primary tumor types regardless of imbalanced representation. Additionally, we adapted the models to take as input either all Ensembl genes (60,483) or protein coding genes only (19,758). Unlike previous work, we avoided selection bias by not filtering genes based on expression values. RNA-seq count data expressed as FPKM-UQ of 9,025 and 10,940 samples from The Cancer Genome Atlas (TCGA) were downloaded from the Genomic Data Commons (GDC) corresponding to 17 and 32 primary tumor types respectively for training and validating the models. Results: All 4 1D-CNN models had an overall accuracy of 94.7% to 97.6% on the test dataset. Further evaluation indicates that the models with protein coding genes only as features performed with better accuracy compared to the models with all Ensembl genes for both 17 and 32 primary tumor types. For all models, the accuracy by primary tumor type was above 80% for most primary tumor types. Conclusions: We packaged all 4 models as a Python-based deep learning classification tool called TULIP (TUmor CLassIfication Predictor) for performing quality control on primary tumor samples and characterizing cancer samples of unknown tumor type. Further optimization of the models is needed to improve the accuracy of certain primary tumor types.

Jones, Sara↗
DOE OSTI2025

Landscaper v1

Understanding the inner workings of machine learning models through their loss landscapes offers crucial insights into model properties, optimization dynamics, and generalizability. However, accessing these insights has traditionally required specialized mathematical expertise, limiting broader adoption. Landscaper is an open-source Python package designed to bridge this gap. Landscaper seamlessly integrates a suite of multi-dimensional loss landscape analyses with cutting-edge topological data analysis (TDA) methods. This powerful combination makes both fundamental loss landscape analysis and advanced TDA techniques accessible to the broader scientific ML community, without requiring deep pre-existing mathematical knowledge. Landscaper offers three key functionalities: * Construction: Builds detailed loss landscape representations through versatile low and high-dimensional sampling techniques. * Quantification: Applies advanced metrics, including a novel topological data analysis (TDA) based smoothness metric, enabling new perspectives on model behavior. * Visualization: Offers intuitive tools to visualize and interpret loss landscapes, providing actionable insights beyond traditional performance metrics.

Weber, Gunther [Lawrence Berkeley National Laborat↗
NASA NTRS1990

An infrared sky model based on the IRAS point source data

A detailed model for the infrared point source sky is presented that comprises geometrically and physically realistic representations of the galactic disk, bulge, spheroid, spiral arms, molecular ring, and absolute magnitudes. The model was guided by a parallel Monte Carlo simulation of the Galaxy. The content of the galactic source table constitutes an excellent match to the 12 micrometer luminosity function in the simulation, as well as the luminosity functions at V and K. Models are given for predicting the density of asteroids to be observed, and the diffuse background radiance of the Zodiacal cloud. The model can be used to predict the character of the point source sky expected for observations from future infrared space experiments.

Cohen, Martin↗
DOE OSTI2022

Multivariate Time Series Anomaly Detection with Few Positive Samples

Given the scarcity of anomalies in real-world applications, the majority of literature has been focusing on modeling normality. The learned representations enable anomaly detection as the normality model is trained to capture certain key underlying data regularities under normal circumstances. In practical settings, particularly industrial time series anomaly detection, we often encounter situations where a large amount of normal operation data is available along with a small number of anomaly events collected over time. This practical situation calls for methodologies to leverage these small number of anomaly events to create a better anomaly detector. In this paper, we introduce two methodologies to address the needs of this practical situation and compared them with recently developed state of the art techniques. Our proposed methods anchor on representative learning of normal operation with autoregressive (AR) model along with loss components to encourage representations that separate normal versus few positive examples. We applied the proposed methods to two industrial anomaly detection datasets and demonstrated effective performance in comparison with approaches from literature. Our study also points out additional challenges with adopting such methods in practical applications.

Xue, Feng↗
NASA NTRSDate not supplied

Evaluating the Global Water and Energy Cycles in Reanalyses Using Ensemble Spread

Reanalyses provide continuous observation based global data over weather and climate time scales through observational assimilation. Ensemble forecasts are a component of modern data assimilation systems, providing information to the analysis process. Storing this spread over the duration of a climate reanalysis will provide insight to the variations of the model that underlies the reanalysis data. Global water and energy cycles (WECs) are a critical aspect of both weather and climate, yet there are some large gaps in the observation of key components as well as their representation in models and reanalyses. NASA’s Water and Energy cycle Studies (NEWS) program is developing a corrected and balanced budget data set that is closed at monthly and regional scales during the 21st century, so far. In this paper, we will evaluate reanalysis WEC components against the initial NEWS collection of observation, but integrate the analysis spread from the reanalysis data assimilation as a new perspective on the evaluation. The initial evaluation will cover a few more recent years using ERA5 10 member spread. As the GMAO’s production of the Modern-Era Retrospective analysis for Research and Applications 21st Century (MERRA-21C) progresses, the effort will include the MERRA-21C 32 member spread. Ultimately, we will explore the potential of the ensemble spread as a measure of uncertainty in the reanalysis.

Michael Bosilovich↗
DOE OSTI2015

Gamification and Visualization of Sensor Data Analysis in Research Buildings

The use of video game elements in non-gaming systems, or gamification , has potential value in transforming data analysis. Our study focused on creating a web-based videogame that models two physical test buildings, each of which contains hundreds of sensors. After the application renders the models, the player can walk through the environments and interact with the virtual representations of the sensors inside. Rather than trudging through a database with textual commands and screens full of data, the user can (virtually) walk up to a sensor and view its data graphically. But these features only scratch the surface of what is possible using our new gamification approach. We anticipate being able to show that recent progress in game design techniques and capacities can contribute to the field of analysis through gamification. The net result could be more stimulating, intuitive, user-friendly interfaces, as well as potentially more informative and insightful applications.

Stone, Jackson A↗
DOE OSTI2015

Thermodynamic assessment of the Pr-O system

We found that the Calphad method was used to perform a thermodynamic assessment of the Pr–O system. Compound energy formalism representations were developed for the fluorite α-PrO 2–x and bixbyite σ-Pr 3 O 5 ± x solid solutions while the two-sublattice liquid model was used to describe the binary melt. The series of phases between Pr 2 O 3 and PrO 2 were taken to be stoichiometric. Moreover, the equilibrium oxygen pressure, phase equilibria, and enthalpy data were used to optimize the adjustable parameters of the models for a self-consistent representation of the thermodynamic behavior of the Pr–O system from 298 K to melting.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS1975

Development of thermal stratification and destratification scaling concepts. Volume 2: Stratification

Temperature and pressure data obtained from the saturated Freon 113 PCA closed-tank stratification tests are presented. The data presented in tabular form are the test conditions, sensible heat values, and Freon 113 PCA liquid and ullage (vapor) properties. Also included, are graphical representations of the liquid bulk temperature and pressure histories, and dimensionless liquid-ullage delta-temperature profiles. Modified Grashof numbers and Fourier number-history data are also presented graphically.

Lovrich, T. N.↗
NASA NTRS2010

Joint Composable Object Model and LVC Methodology

Within the Department of Defense, multiple architectures are created to serve and fulfill one or several specific service or mission related LVC training goals. Multiple Object Models exist across and within those architectures and it is there that those disparate object models are a major source of interoperability problems when developing and constructing the training scenarios. The two most commonly used architectures are; HLA and TENA, with DIS and CTIA following close behind in terms of the number of users. Although these multiple architectures can share and exchange data the underlying meta-models for runtime data exchange are quite different, requiring gateways/translators to bridge between the different object model representations; while the Department of Defense's use of gateways are generally effective in performing these functions, as the LVC environment increases so too does the cost and complexity of these gateways. Coupled with the wide range of different object models across the various user communities we increase the propensity for run time errors, increased programmer stop gap measures during coordinated exercises, or failure of the system as a whole due to unknown or unforeseen incompatibilities. The Joint Composable Object Model (JCOM) project was established under an M&S Steering Committee (MSSC)-sponsored effort with oversight and control placed under the Joint Forces Command J7 Advanced Concepts Program Directorate. The purpose of this paper is to address the initial and the current progress that has been made in the following areas; the Conceptual Model Development Format, the Common Object Model, the Architecture Neutral Data Exchange Model (ANDEM), and the association methodology to allow the re-use of multiple architecture object models and the development of the prototype persistent reusable library.

Rheinsmith, Richard↗
NASA NTRS1985

Mars atmospheric circulation - Aspects from Viking Landers

Winds measured by the two Viking Landers have been filtered and then compared with predictions from the general circulation model and to Orbiter observations of clouds and surface phenomena that indicate wind direction. This was done to determine the degree to which filtered winds may represent aspects of the general circulation. Excellent agreement was found between wind direction data from Lander 1 and the model predictions and Orbiter observations. For Lander 2, agreement was generally good, but there were periods of disagreement which indicate that the filtering did not remove other extraneous effects. It is concluded that Lander 1 gives a good representation of the general circulation at 22.5 deg N latitude but that Lander 2 is suspect. Most wind data from Lander 1 have yet to be analyzed. It appears that when analyzed these Lander 1 data (covering 3.5 Mars years) can provide information about interannual variations in the general circulation at the Lander latitude.

Ryan, J. A.↗
DOE OSTI2025

Spectral Properties of Globally Distributed ENA Fluxes across Diverse Regions of the Heliosphere

This study analyzes energetic neutral atom (ENA) spectral properties across distinct regions of globally distributed flux (GDF) sky maps, using Interstellar Boundary Explorer data from a full solar cycle, corrected for time dispersion. By time-shifting the data to the heliosheath using GDF source distances from D. B. Reisenfeld et al., we achieve a more accurate representation of heliosheath GDF energy spectra. We quantify ENA spectral characteristics, heliosheath line-of-sight-integrated proton pressure, and heliosheath proton temperature, comparing these to solar wind properties at 1 au and interplanetary scintillation-derived solar wind data. Our findings show that the spectral index is generally anticorrelated with heliosheath proton temperature and pressure, except in the central tail, where a partial positive correlation is observed. The lowest spectral index values occur when high-latitude heliosheath regions are dominated by fast solar wind from polar coronal holes. The south pole exhibits the flattest energy spectra due to plasma heating from both fast solar wind and a late-2014 pressure pulse. The central tail shows shorter variability (5–6 yr) for spectral index and heliosheath proton temperature, while proton pressure follows the 11 yr solar cycle. Most spectral shapes exhibit a “knee” distribution, peaking during solar maximum, with an “ankle” shape observed only at the south pole during solar cycle transitions. Asymmetry in proton pressure in the lobes is driven by the draping effect of the local interstellar magnetic field. This study provides insights into the energetic properties of GDF across the heliosphere, enhancing our understanding of the heliospheric environment.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2017

Improvements in Electron-Photon-Relaxation Data for MCNP6

The Los Alamos Monte Carlo transport code MCNP6 has recently benefitted from significant work to improve the photon-electron transport and atomic-relaxation capabilities of the code. The improvements include extension of existing photon data to lower energies than previously supported; addition of new data such as subshell-specific photoelectric cross sections; completion of the incoherent and coherent form factor data to cover all scattering angles at all supported energies; new data for electron interactions including Bremsstrahlung, elastic scattering, subshell-specific electroionization, and atomic excitation; a much more detailed representation for atomic relaxation; and an entirely new (to MCNP) approach to electron transport: the single-event method. These improvements were made available with the initial production release of MCNP6 along with an extended electron, photon, and relaxation data file eprdata12 to support the new capabilities. More recently new data have become available and have now been incorporated into MCNP6, correcting a few issues with the previous work, and adding new sets of cross sections that were not included in eprdata12. The purpose of this paper is to introduce these newer changes and enhancements, which will be included with the next public release of the MCNP6 code system. The database that supports these developments relies on an evolving set of libraries originally released by D. E. Cullen and collaborators as separate data for photo-atomic interactions (EPDL), atomic relaxation (EADL), and electro-atomic interactions (EEDL). An updated version of this release was officially adopted in the ENDF/B VI.8 release of the Evaluated Nuclear Data File and it was this source that was used to produce the ACE (A Compact ENDF) format library eprdata12 for use in the initial production release of MCNP6.1 (and the subsequent interim release MCNP6.1.1). The latest improved version of this database was released by Cullen in 2014. This is the EPICS2014 collection, which includes both an ENDF-formatted file and separate EPDL, EADL, and EEDL files. In the area of photon-electron transport and atomic-relaxation, the newest enhancements to the current version of MCNP6 (scheduled for release as MCNP6.2) are primarily based on improvements and additions drawn from the EPICS2014 data.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS1993

Identification of modeled ocean plumes in Greenland gyre ERS-1 SAR data

Oceanic convective plumes modeled with a thermobaric large-eddy simulation and driven by conditions similar to those of the Greenland Sea are compared to observations from ERS-1 Synthetic Aperture Radar (SAR) data from the Greenland Sea for the winter of 1992. In both form and size the two representations are seen to compare favorably. The plume-filled area of the SAR image occupies a region about 20 km by 90 km at the ice edge of the open water in 'Nordbukta', the large seasonal ice retreat, in the 'Odden' ice protuberance in the southern Greenland gyre. In the SAR data the plumes appear to be ice covered while the convective-return areas are open.

Carsey, Frank D.↗