Search NASA⌕ Search

SEARCH · Search NASA

Results for “test vector”

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 343 records · Page 19
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Overview of new MAST physics in anticipation of first results from MAST Upgrade

We report that the mega amp spherical tokamak (MAST) was a low aspect ratio device (R/a = 0.85/0.65 ~ 1.3) with similar poloidal cross-section to other medium-size tokamaks. The physics programme concentrates on addressing key physics issues for the operation of ITER, design of DEMO and future spherical tokamaks by utilising high resolution diagnostic measurements closely coupled with theory and modelling to significantly advance our understanding. An empirical scaling of the energy confinement time that favours higher power, lower collisionality devices is consistent with gyrokinetic modelling of electron scale turbulence. Measurements of ion scale turbulence with beam emission spectroscopy and gyrokinetic modelling in up-down symmetric plasmas find that the symmetry of the turbulence is broken by flow shear. Near the non-linear stability threshold, flow shear tilts the density fluctuation correlation function and skews the fluctuation amplitude distribution. Results from fast particle physics studies include the observation that sawteeth are found to redistribute passing and trapped fast particles injected from neutral beam injectors in equal measure, suggesting that resonances between the m = 1 perturbation and the fast ion orbits may be playing a dominant role in the fast ion transport. Measured D–D fusion products from a neutron camera and a charged fusion product detector are 40% lower than predictions from TRANSP/NUBEAM, highlighting possible deficiencies in the guiding centre approximation. Modelling of fast ion losses in the presence of resonant magnetic perturbations (RMPs) can reproduce trends observed in experiments when the plasma response and charge-exchange losses are accounted for. Measurements with a neutral particle analyser during merging-compression start-up indicate the acceleration of ions and electrons. Transport at the plasma edge has been improved through reciprocating probe measurements that have characterised a geodesic acoustic mode at the edge of an ohmic L-mode plasma and particle-in-cell modelling has improved the interpretation of plasma potential estimates from ball-pen probes. The application of RMPs leads to a reduction in particle confinement in L-mode and H-mode and an increase in the core ionization source. The ejection of secondary filaments following type-I ELMs correlates with interactions with surfaces near the X-point. Simulations of the interaction between pairs of filaments in the scrape-off layer suggest this results in modest changes to their velocity, and in most cases can be treated as moving independently. A stochastic model of scrape-off layer profile formation based on the superposition of non-interacting filaments is in good agreement with measured time-average profiles. Transport in the divertor has been improved through fast camera imaging, indicating the presence of a quiescent region devoid of filament near the X-point, extending from the separatrix to $ψ_n$ ~ 1.02. Simulations of turbulent transport in the divertor show that the angle between the divertor leg on the curvature vector strongly influences transport into the private flux region via the interchange mechanism. Coherence imaging measurements show counter-streaming flows of impurities due to gas puffing increasing the pressure on field lines where the gas is ionised. MAST Upgrade is based on the original MAST device, with substantially improved capabilities to operate with a Super-X divertor to test extended divertor leg concepts. SOLPS-ITER modelling predicts the detachment threshold will be reduced by more than a factor of 2, in terms of upstream density, in the Super-X compared with a conventional configuration and that the radiation front movement is passively stabilised before it reaches the X-point. 1D fluid modelling reveals the key role of momentum and power loss mechanisms in governing detachment onset and evolution. Analytic modelling indicates that long legs placed at large major radius, or equivalently low at the target compared with the X-point are more amenable to external control. With MAST Upgrade experiments expected in 2019, a thorough characterisation of the sources of the intrinsic error field has been carried out and a mitigation strategy developed.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Scalable learning of potentials to predict time-dependent Hartree–Fock dynamics

We propose a framework to learn the time-dependent Hartree–Fock (TDHF) inter-electronic potential of a molecule from its electron density dynamics. Although the entire TDHF Hamiltonian, including the inter-electronic potential, can be computed from first principles, we use this problem as a testbed to develop strategies that can be applied to learn a priori unknown terms that arise in other methods/approaches to quantum dynamics, e.g., emerging problems such as learning exchange–correlation potentials for time-dependent density functional theory. We develop, train, and test three models of the TDHF inter-electronic potential, each parameterized by a four-index tensor of size up to 60 × 60 × 60 × 60. Two of the models preserve Hermitian symmetry, while one model preserves an eight-fold permutation symmetry that implies Hermitian symmetry. Across seven different molecular systems, we find that accounting for the deeper eight-fold symmetry leads to the best-performing model across three metrics: training efficiency, test set predictive power, and direct comparison of true and learned inter-electronic potentials. All three models, when trained on ensembles of field-free trajectories, generate accurate electron dynamics predictions even in a field-on regime that lies outside the training set. To enable our models to scale to large molecular systems, we derive expressions for Jacobian-vector products that enable iterative, matrix-free training.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

New thermal-relic targets for sub-GeV dark matter direct detection

Dark matter direct detection experiments involving electron recoils are beginning to test highly predictive, thermal-relic milestones for sub-GeV dark matter models. Due to the Lee-Weinberg bound, thermal dark matter candidates in this mass range necessarily require comparably light mediator particles to achieve a suitably large annihilation cross section. Here we present new thermal-relic milestones for sub-GeV dark matter candidates that couple to vector mediators. In these models, the mediators are massive gauge bosons of anomaly-free Abelian extensions to the Standard Model, including the dark photon, gauged L i - L j , B - L , and B - 3 L i models, where B is the baryon number, L is the lepton number, and i , j index the lepton families. Since the same interactions that govern cosmological production also govern electron scattering, the targets we present are firmly predictive and allow for these models to be robustly discovered or falsified. Furthermore, since the mediators we study exhaust the minimal anomaly-free U(1) extensions to the Standard Model, our results offer a complete list of predictive milestones for sub-GeV dark matter coupled to vector mediators.

Han, Xu [Chicago U.] (ORCID:0009000821607836)↗
DOE OSTI2025

Real-World Cyber Security Demonstration for Networked Electric Drives

In this article, we present the design and implementation of a cyber-physical security testbed for networked electric drive systems, aimed at conducting real-world security demonstrations. To our knowledge, this is one of the first security testbeds for networked electric drives, seamlessly integrating the domains of power electronics and computer science, and cybersecurity. By doing so, the testbed offers a comprehensive platform to explore and understand the intricate and often complex interactions between cyber and physical systems. The core of our testbed consists of four electric machine drives, meticulously configured to emulate small-scale but realistic information technology (IT) and operational technology (OT) networks. This setup both provides a controlled environment for simulating a wide array of cyber-attacks, and mirrors potential real-world attack scenarios with a high degree of fidelity. The testbed serves as an invaluable resource for the study of cyber-physical security, offering a practical and dynamic platform for testing and validating cybersecurity measures in the context of networked electric drive systems. As a concrete example of the testbed's capabilities, we have developed and implemented a Python-based script designed to execute step-stone attacks over a wireless local area network (WLAN). This script leverages a sequence of target IP addresses, simulating a real-world attack vector that could be exploited by adversaries. To counteract such threats, we demonstrate the efficacy of our developed cyber-attack detection algorithms, which are integral to our testbed's security framework. Furthermore, the testbed incorporates a real-time visualization system using InfluxDB and Grafana, providing a dynamic and interactive representation of networked electric drives and their associated security monitoring mechanisms. This visualization component not only enhances the testbed's usability but also offers insightful, real-time data for researchers and practitioners, thereby facilitating a deeper understanding of cyber-physical security dynamics in networked electric drive systems.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2019

Protein expression and gene editing in monocots using foxtail mosaic virus vectors

Abstract Plant viruses can be engineered to carry sequences that direct silencing of target host genes, expression of heterologous proteins, or editing of host genes. A set of foxtail mosaic virus (FoMV) vectors was developed that can be used for transient gene expression and single guide RNA delivery for Cas9‐mediated gene editing in maize, Setaria viridis , and Nicotiana benthamiana . This was accomplished by duplicating the FoMV capsid protein subgenomic promoter, abolishing the unnecessary open reading frame 5A, and inserting a cloning site containing unique restriction endonuclease cleavage sites immediately after the duplicated promoter. The modified FoMV vectors transiently expressed green fluorescent protein (GFP) and bialaphos resistance (BAR) protein in leaves of systemically infected maize seedlings. GFP was detected in epidermal and mesophyll cells by epifluorescence microscopy, and expression was confirmed by Western blot analyses. Plants infected with FoMV carrying the bar gene were temporarily protected from a glufosinate herbicide, and expression was confirmed using a rapid antibody‐based BAR strip test. Expression of these proteins was stabilized by nucleotide substitutions in the sequence of the duplicated promoter region. Single guide RNAs expressed from the duplicated promoter mediated edits in the N. benthamiana Phytoene desaturase gene, the S. viridis Carbonic anhydrase 2 gene, and the maize HKT1 gene encoding a potassium transporter. The efficiency of editing was enhanced in the presence of synergistic viruses and a viral silencing suppressor. This work expands the utility of FoMV for virus‐induced gene silencing (VIGS), virus‐mediated overexpression (VOX), and virus‐enabled gene editing (VEdGE) in monocots.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Understanding and quantifying electron beam effects during in situ TEM nanomechanical tensile testing on metal thin films

Transmission electron microscopy (TEM) imaging relies on high energy electrons for atomic scale resolution, however, the electrons themselves interact with and may alter the material being imaged. Using an in situ TEM MEMS-based nanomechanical testing technique, the effect of the electron beam (e-beam) on the deformation behavior of nanocrystalline Al and ultrafine-grained Au is investigated and quantified. We show that the e-beam enhances plastic deformation, leading to an increase in plastic strain rate and a decrease in true activation volume V* in Al (28 to 21b 3 , with b being the Burgers vector length). The e-beam has a much weaker effect on Au. The e-beam effect is not caused by knock-on damage, but rather an effective temperature increase due to additional atomic fluctuations provided by the e-beam. The effective temperature increase is larger for Al than Au. This e-beam effect does not change the deformation mechanisms, but instead accelerates the stress-driven, thermally activated plastic deformation. These experiments provide insight into the effects of the e-beam on plastic deformation in different metals and underscore the importance of understanding and quantifying these effects for proper interpretation of measured mechanical properties during in situ TEM experiments.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Deep Learning Estimation of Daily Ground–Level NO 2 Concentrations from Remote Sensing Data

The limited number of nitrogen dioxide (NO 2 ) surface measurements calls for the development of highly accurate approaches to estimating surface NO 2 concentrations. In this study, we leverage a new satellite instrument, the TROPOspheric Monitoring Instrument (TROPOMI), along with other predictor variables, to estimate daily surface NO 2 concentrations over Texas in 2019. We use the deep convolutional neural network (Deep-CNN), an advanced deep learning algorithm, to obtain estimates and achieve a correlation coefficient (R) of 0.91, an index of agreement (IOA) of 0.95, and a mean absolute bias (MAB) of 1.75 ppb in surface NO 2 estimation. Additionally, we leverage a novel approach, SHapley Additive exPlanations (SHAP), to describe how Deep-CNN understands each predictor variable. The SHAP results show that the Deep-CNN model has an advanced understanding of the dataset, revealing that TROPOMI closely captures levels of NO 2 . In addition, we show the superiority of our Deep-CNN model at estimating surface NO 2 over other well-known machine learning and regression models in the field, including the support vector machines (SVM), random forest (RF), and multiple linear regression (MLR). Although SVM and RF show strong capabilities at estimating surface NO 2 concentrations, their accuracy is inferior to that of the Deep-CNN model, ranking second and third in model accuracy in this study. The MLR, however, shows a poor ability at NO 2 estimation and ranks last among all models. Furthermore, testing the impact of sample size on model performance, we also show that, compared to other models, Deep-CNN needs more samples to trigger its strength at surface NO 2 estimation.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Quantitative x-ray scattering of free molecules

Advances in x-ray free electron lasers have made ultrafast scattering a powerful method for investigating molecular reaction kinetics and dynamics. Accurate measurement of the ground-state, static scattering signals of the reacting molecules is pivotal for these pump-probe x-ray scattering experiments as they are the cornerstone for interpreting the observed structural dynamics. Here, this article presents a data calibration procedure, designed for gas-phase x-ray scattering experiments conducted at the Linac Coherent Light Source x-ray Free-Electron Laser at SLAC National Accelerator Laboratory, that makes it possible to derive a quantitative dependence of the scattering signal on the scattering vector. A self-calibration algorithm that optimizes the detector position without reference to a computed pattern is introduced. Angle-of-scattering corrections that account for several small experimental non-idealities are reported. Their implementation leads to near quantitative agreement with theoretical scattering patterns calculated with ab-initio methods as illustrated for two x-ray photon energies and several molecular test systems.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2022

First Leptophobic Dark Matter Search from the Coherent–CAPTAIN-Mills Liquid Argon Detector

We report the first results of a search for leptophobic dark matter (DM) from the Coherent–CAPTAIN-Mills (CCM) liquid argon (LAr) detector. An engineering run with 120 photomultiplier tubes (PMTs) and 17.9 × 10 20 protons on target (POT) was performed in fall 2019 to study the characteristics of the CCM detector. The operation of this 10-ton detector was strictly light based with a threshold of 50 keV and used coherent elastic scattering off argon nuclei to detect DM. Despite only 1.5 months of accumulated luminosity, contaminated LAr, and nonoptimized shielding, CCM’s first engineering run has already achieved sensitivity to previously unexplored parameter space of light dark matter models with a baryonic vector portal. With an expected background of 115 005 events, we observe 115 005 + 16.5 events which is compatible with background expectations. For a benchmark mediator-to-DM mass ratio of m V B =m χ = 2.1, DM masses within the range 9 MeV ≲ m χ ≲ 50 MeV are excluded at 90% C. L. in the leptophobic model after applying the Feldman-Cousins test statistic. CCM’s upgraded run with 200 PMTs, filtered LAr, improved shielding, and 10 times more POT will be able to exclude the remaining thermal relic density parameter space of this model, as well as probe new parameter space of other leptophobic DM models.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Muonic force behind flavor anomalies

We develop an economical theoretical framework for combined explanations of the flavor physics anomalies involving muons: (g - 2) μ , R K(*), and b → sμ + μ - angular distributions and branching ratios, that was first initiated by some of us in ref. [1]. The Standard Model (SM) is supplemented with a lepton-flavored U(1) X gauge group. The U(1) X gauge boson with the mass of O(0.1) GeV resolves the (g - 2) μ tension. A TeV-scale leptoquark, charged under the U(1) X , carries a muon number and mediates B-decays without prompting charged lepton flavor violation or inducing proton decay. We explore the theory space of the chiral, anomaly-free U(1) X gauge extensions featuring the above scenario, and identify many suitable charge assignments for the SM+3ν R fermion content with the integer charges in the range X F i $\in$ [-10, 10]. We then carry out a comprehensive phenomenological study of the muonic force in representative benchmark models. Interestingly, we found models which can resolve the tension without conflicting the complementary constraints, and all of the viable parameter space will be tested in future muonic resonance searches. Finally, the catalog of the anomaly-free lepton-non-universal charge assignments motivated us to explore different directions in model building. We present a model in which the muon mass and the (g - 2) μ are generated radiatively from a common short-distance dynamics after the U(1) X breaking. We also show how to charge a vector leptoquark under U(1) μ-τ in a complete gauge model.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Control Strategies and Validation in the Hybrid Optimization and Performance Platform (HOPP)

The Hybrid Optimization and Performance Platform (HOPP) is a tool that simulates hybrid power plants in various configurations, and also calculates the financial feasibility of these plants. This report outlines an overview of HOPP and the energy storage dispatch strategies available. It then presents three case studies which demonstrate different applications of HOPP. The first case looks at the profitability of hybrid power plants in different locations in the USA. The second case examines the availability of hybrid power plants to provide energy reliability services. The third case presents a plant that produces both hydrogen and electricity, and demonstrates a dispatch strategy that chooses the most profitable energy vector based on price signals. The next section shows the validation of HOPP on operational data, using data from both unit-scale and utility-scale power plants. This validation process demonstrated that HOPP can simulate the power output of both wind and solar PV plants at both scales with comparable fidelity to an existing commercial software tool. Finally, HOPP is applied in a field test which applies an optimal dispatch strategy to a physical battery in a unit-scale hybrid plant at NREL. HOPP's optimal dispatch strategy, applied in a real-world setting, improved this hybrid plant's ability to meet a load signal while minimizing operational costs.

14 SOLAR ENERGY↗
DOE OSTI2024

High-Throughput Screening and Accurate Prediction of Ionic Liquid Viscosities Using Interpretable Machine Learning

Ionic liquids (ILs) are a novel group of green solvents with great promise for various industrial applications, including carbon capture and lignocellulosic biomass deconstruction. However, the use of ILs at the industrial scale remains challenging due to their high viscosities at ambient temperatures. To develop ILs with lower viscosities, a systematic study of their quantitative structure–property relationship (QSPR) is desirable. Here, we developed four machine learning (ML) models to predict viscosity at various temperature and pressure ranges, trained over a wide range of ILs consisting of various cationic and anionic families. ML methods including two-factor polynomial regression (two-factor PR), support vector regression (SVR), feed-forward neural networks (FFNN), and categorical boosting (CATBoost) were developed based on features that have proven useful in previous ML studies: COSMO-RS (conductor-like screening model for real solvents)-derived surface screening charge densities (sigma profiles). FFNN and CATBoost were the most accurate in predicting IL viscosities with lower average absolute relative deviation and higher R2 values on the test set. Tanimoto similarity scores were calculated to characterize the chemical space and structural similarity of the investigated ions. Furthermore, SHapley Additive exPlanation (SHAP) analysis was employed to interpret the ML results. Temperature, the polar area of ILs, and the nonpolar regions of ions are key features that influence the viscosity predictions. Importantly, the IL viscosity prediction here is the most accurate reported to date.

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

A three-point velocity estimation method for two-dimensional coarse-grained imaging data

Time delay and velocity estimation methods have been widely studied subjects in the context of signal processing, with applications in many different fields of physics. The velocity of waves or coherent fluctuation structures is commonly estimated as the distance between two measurement points divided by the time lag that maximizes the cross correlation function between the measured signals, but this is demonstrated to result in erroneous estimates for two spatial dimensions. We present an improved method to accurately estimate both components of the velocity vector, relying on three non-aligned measurement points. We introduce a stochastic process describing the fluctuations as a superposition of uncorrelated pulses moving in two dimensions. Using this model, we show that the three-point velocity estimation method, using time delays calculated through cross correlations, yields the exact velocity components when all pulses have the same velocity. The two- and three-point methods are tested on synthetic data generated from realizations of such processes for which the underlying velocity components are known. The results reveal the superiority of the three-point technique. Finally, we demonstrate the applicability of the velocity estimation on gas puff imaging data of strongly intermittent plasma fluctuations due to the radial motion of coherent, blob-like structures at the boundary of the Alcator C-Mod tokamak.

Materials Science↗
DOE OSTI2023

Using optimal estimation to retrieve winds from velocity-azimuth display (VAD) scans by a Doppler lidar

Abstract. Low-powered commercially available coherent Doppler lidar (CDL) wind profilers provide continuous measurement of vertical profiles of wind in the lower troposphere, usually close to or up to the top of the planetary boundary layer. The vertical extent of these wind profiles is limited by the availability of scatterers and thus varies substantially throughout the day and from one day to the next. This makes it challenging to develop continuous products that rely on CDL-observed wind profiles. In order to overcome this problem, we have developed a new method for wind profile retrievals from CDL that combines the traditional velocity-azimuth display (VAD) technique with optimal estimation (OE) to provide continuous wind profiles up to 3 km. The new method exploits the level-to-level covariance present in the wind profile to fill in the gaps where the signal-to-noise ratio of the CDL return is too low to provide reliable results using the traditional VAD method. Another advantage of the new method is that it provides the full error covariance matrix of the solution and profiles of information content, which more easily facilitates the assimilation of the observed wind profiles into numerical weather prediction models. This method was tested using yearlong CDL measurements at the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) Central Facility in 2019. Comparison with the ARM operational CDL wind profile product and collocated radiosonde wind measurements shows excellent agreement (R2>0.99) with no degradation in results where the traditional VAD provided a valid solution. In the region where traditional VAD does not provide results, the OE wind speed and wind vector have uncertainties of 3.44 and 4.33 m s−1, respectively. As a result, the new method provides additional information over the standard technique and increases the effective range of existing CDL systems without the need for additional hardware.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2019

A new high-order spectral difference method for simulating viscous flows on unstructured grids with mixed-element meshes

In the present study, a spectral difference (SD) method is developed for viscous flows on meshes with a mixture of triangular and quadrilateral elements. The standard SD method for triangular elements, which employs Lagrangian interpolating functions for fluxes, is not stable when the designed accuracy of spatial discretization is third order or higher. Unlike the standard SD method, the method examined here uses vector interpolating functions in the Raviart-Thomas (RT) spaces to construct continuous flux functions on reference elements. The spectral-difference Raviart-Thomas (SDRT) method was originally proposed by Balan et al. and implemented on triangular-element meshes for invisid flow only. Our present results demonstrated that the SDRT method is stable and high-order accurate in two dimensions (2D) for a number of test problems by using triangular-, quadrilateral-, and mixed-element meshes for both inviscid and viscous flows. A stability analysis is also performed for mixed elements. Furthermore, we found our current SDRT scheme is stable for fourth-order accurate spatial discretizations. However, a stable fifth-order SDRT scheme remains to be identified.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Automated Stellar Spectra Classification with Ensemble Convolutional Neural Network

Large sky survey telescopes have produced a tremendous amount of astronomical data, including spectra. Machine learning methods must be employed to automatically process the spectral data obtained by these telescopes. Classification of stellar spectra by applying deep learning is an important research direction for the automatic classification of high-dimensional celestial spectra. In this paper, a robust ensemble convolutional neural network (ECNN) was designed and applied to improve the classification accuracy of massive stellar spectra from the Sloan digital sky survey. We designed six classifiers which consist six different convolutional neural networks (CNN), respectively, to recognize the spectra in DR16. Then, according the cross-entropy testing error of the spectra at different signal-to-noise ratios, we integrate the results of different classifiers in an ensemble learning way to improve the effect of classification. The experimental result proved that our one-dimensional ECNN strategy could achieve 95.0% accuracy in the classification task of the stellar spectra, a level of accuracy that exceeds that of the classical principal component analysis and support vector machine model.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Machine Learned Empirical Numerical Integrator from Simulated Data

Recently, a number of state-of-the-art surrogate machine learning (ML) models have been designed for global weather and climate prediction, which have been trained using reanalysis data products. Reanalysis data products are constructed using numerical model simulations that combine numerical integration of partial differential equations and parameterization schemes. These products are typically only archived and made available using coarsened spatial and temporal resolutions. This study explores the impact of the numerical generation methods used to produce the training datasets and the temporal resolution of those datasets on machine learning surrogate models. Using the nonlinear vector autoregression (NVAR) machine as an explainable ML technique, simple dynamical systems are emulated with ML models trained on data produced by three classical numerical integration schemes. NVAR is validated as a skillful ML method, capable of producing accurate predictions and, more importantly, reconstructing both the underlying dynamics and the numerical integration scheme used to generate the training data. However, the machine fails to generalize predictions on unseen test data generated by different numerical integration schemes, despite the underlying dynamical system being the same. This result provides a word of caution for the growing field of machine learning emulation of weather and climate dynamics. Furthermore, we illustrate using NVAR that training on temporally coarsened data may increase the required complexity of ML models and potentially introduce new numerical challenges. Finally, we discover that empirical integration schemes with arbitrary time-stepping sizes can be constructed directly from the data, which implies a potential for the development of empirical numerical integration schemes.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Attention-Augmented Parametric Kernel Graph Neural Network (APKGNN) for Node Classification

We present a new graph neural network, the Attention-based Parametric-Kernel augmented Graph Neural Network (APKGNN), developed for node classification tasks. Despite extensive work on modeling multi-faceted relationships between connected nodes of a graph, the effect of attention on edge features mapped to relationships has not yet been analyzed through learning representation. This study derives such an attention vector by first calculating node features corresponding to endpoints of an edge and then aggregating these with extracted local intrinsic patches of a given graph to generate augmented local patch vectors. This process uses a parametric kernel based on Gaussian mixture models (GMMs) to embed local neighborhoods of the graph in local patches. The patch vectors then convolve with the above node features to produce an updated node representation. We show that this new learning representation (APKGNN) achieves higher node classification accuracy on tasks - both standard benchmarks (Cora, PubMed, Citeseer) and new experimental short text corpora where nodes correspond to text documents and words. This implementation of the GNN convolution layer outperforms state-of-the-art (SOTA) algorithms, achieving higher training, validation, and test accuracy by a significant margin on three standard benchmark data sets under both SOTA experimental settings and those for new testbeds.

Bose, Avishek↗