Search NASA⌕ Search

SEARCH · Search NASA

Results for “Gaussian Process 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 397 records · Page 22

Activity coefficient acquisition with thermodynamics–informed active learning for phase diagram construction

This work explores the use of thermodynamics-informed Gaussian processes (GPs) and active learning (AL) to model activity coefficients and construct phase diagrams. Relying on synthetic data generated from an excess Gibbs energy model, GPs were found to accurately describe the activity coefficients of several binary mixtures across large composition and temperature ranges. Moreover, GPs could estimate their own uncertainty and identify composition/temperature regions where activity coefficient data provide the most information to the models. This was leveraged to build AL algorithms targeted at modeling phase equilibria. In many cases, a single active-learning-acquired data point was sufficient to describe the phase diagrams studied. Lastly, the ability of AL to greatly reduce the amount of data needed to obtain accurate models was further verified on experimental case studies, namely individual ion activity coefficients, the solid–liquid and vapor–liquid equilibrium of deep eutectic solvents, and phase equilibria in ternary mixtures.

25 ENERGY STORAGE↗

Decentralized Voltage Control of Large-Scale Distribution System with PVs Based on MADRL

This paper proposes a model-free decentralized control framework for the voltage regulation of large-scale distribution systems through the coordinated control of PV inverters. This is achieved by developing a novel interaction mechanism between the surrogate model and the centralized training and decentralized execution multiagent deep reinforcement learning framework. Specifically, the sparse Gaussian processes regression method is first utilized to develop the surrogate model of the original distribution system for reward calculation during the training stage, where each agent represents a sub-region in the centralized fashion for coordination strategy learning. After that, the learned control rules are used to inform controllers within each sub-region for real-time decisions with only local measurements. Comparative tests among various methods on the EPRI Ckt5 test system demonstrate the effectiveness of the proposed method.

distribution system↗

THE BIMODAL STRUCTURE OF THE SOLAR CYCLE

Some properties of the 11 yr solar cycle can be explained by the current solar dynamo models. However, some other features remain not well understood such as the asymmetry of the cycle, the double-peaked structure, and the “Waldmeier effect” that a stronger cycle tends to have less rise time and a shorter cycle length. We speculate that the solar cycle is governed by a bi-dynamo model forming two stochastic processes depicted by a bimodal Gaussian function with a time gap of about 2 yr, from which the above features can be reasonably explained. The first one describes the main properties of the cycle dominated by the current solar dynamo models, and the second one occurs either in the rising phase as a short weak explosive perturbation or in the declining phase as a long stochastic perturbation. The above function is the best one selected from several in terms of the Akaike information criterion. Through analyzing different distributions, one might speculate about the dominant physical process inside the convection zone. The secondary (main) process is found to be closely associated with complicated (simple) active ranges. In effect, the bi-dynamo model is a reduced form of a multi-dynamo model, which could occur from the base of the convection zone through its envelope and from low to high heliographic latitude, reflecting the active belts in the convection zone. These results are insensitive to the hemispheric asymmetry, smoothing filters, and distribution functions selected and are expected to be helpful in understanding the formation of solar and stellar cycles.

79 ASTRONOMY AND ASTROPHYSICS↗

Surrogate models for mixed discrete-continuous variables

Large-scale computational models have become common tools for analyzing complex man made systems. However, when coupled with optimization or uncertainty quantification methods in order to conduct extensive model exploration and analysis, the computational expense quickly becomes intractable. Furthermore, these models may have both continuous and discrete parame ters. One common approach to mitigating the computational expense is the use of response surface approximations. While well developed for models with continuous parameters, they are still new and largely untested for models with both continuous and discrete parameters. In this work, we describe and investigate the performance of three types of response surfaces developed for mixed variable models: Adaptive Component Selection and Shrinkage Operator, Treed Gaussian Process, and Gaussian Process with Special Correlation Functions. We focus our efforts on test problems with a small number of parameters of interest, a characteristic of many physics-based engineering models. We present the results of our studies and offer some insights regarding the performance of each response surface approximation method.

Swiler, Laura Painton↗

Sub-pilot-scale Production of High-Value Products from U.S. Coals

Investigators from the University of Utah, University of Wyoming and Marshall University pursued a program to study the conversion of raw coal to high-value products of carbon fiber and silicon carbide. Team members also developed an initial framework for a data portal that can incorporate laboratory data on coal processing and product quality, and also work with tools for machine learning for data analysis, data visualization and economic assessment. Experimental R&D efforts focused on the conversion of raw coal to coal tar and other byproducts, and the resulting tar intermediates were upgraded to form anisotropic and isotropic pitch materials. These pitch materials were produced from coal using both thermal (pyrolysis) and chemical (mild solvolysis liquefaction) decomposition of raw coal. Four different coals were studied: Utah bituminous coal (Sufco), Wyoming PRB coal (Black Thunder), Illinois bituminous coal (Illinois #6), and West Virginia bituminous coal (Flying Eagle). Both metallurgical-grade coking coals and lower-grade steam coals were investigated, and controlled secondary gas-phase reactions were used during a two-stage pyrolysis process to induce cracking and condensation reactions among the pyrolytic tar species. This approach successfully improved the performance of the lower grade coals for yielding pitch materials, with properties more consistent with a commercial-grade pitch that had previously demonstrated success for quality carbon fiber production. The use of waste plastic materials was also studied, to help improve physical and chemical characteristics of the intermediate tars and final pitch product; in particular, for lowering the pitch softening point to an acceptable level for melt spinning carbon fiber. Mild solvolysis liquefaction was also used as a method for producing pitch for carbon fiber production. As expected, significantly higher pitch yields were obtained using this approach, and waste plastic materials were also successfully used to reduce pitch softening point to an acceptable level. The plastic materials were also utilized to create a solvent for the mild solvolysis process, and this plastic-derived solvent was shown to provide results consistent with more expensive commercial chemical solvents, and could thus avoid the need for costly recovery and recycle of a liquefaction solvent. Additional experimental R&D focused on the production of silicon carbide (β-SiC) from the residual char byproduct from pitch production, and also on the production of carbon fiber from the anisotropic pitch. SiC was successfully synthesized using a mixture of residual char and sandstone at a ratio of 1:1. Reaction temperature and residence time were optimized and yielded a product purity of 81%. For carbon fiber production, the most successful pitch samples were obtained from the mild solvolysis liquefaction approach, combined with the use of a plastic (HDPE)-derived solvent. Fiber properties improved over time as laboratory fiber production methodologies improved, and final yields of carbon fiber were obtained with a diameter of 12.14 ± 1.10 um, Modulus of 173.73 ± 15.25 GPa, and Tensile Strength of 1.04 ± 0.10 GPa. A proof-of-concept Modern Community Research Data Portal (MCRDP) was developed and deployed for coal and coal-derived pitch characterization, with the full support of (i) remote web-based access, (ii) distributed analysis, (iii) interactive visualization and exploration, (iv) shared and long-term data access, (v) advanced query capabilities and (vi) real-time collaboration. The Coal to Products Data Portal “coaltoproducts.org” provides researchers with space to store and share data within a project, tools for analyzing and understanding data for scientific investigation, and the ability to publish data to the broader community for reproducibility. The portal leverages the Material Commons 2.0 (MC) platform developed by the Center for PRedictive Integrated Structural Materials Science (PRISMS) of the University of Michigan, to achieve long-term longevity of data collections and, more importantly, collaborative science. A number of data visualization tools were also assessed and implemented for interrogating the experimental and modeling data. The machine learning portion of this project analyzed datasets from two different coal conversion processes performed on a diverse set of coal samples from both the coal pyrolysis experiments and the solvent liquefaction experiments. The work was initiated by exploring standard regression models on the pyrolysis data, aiming to understand the impact of sample characteristics and processing conditions on key product metrics. Over the course of the project, the focus expanded to include a variety of machine learning tools, delving into both supervised and unsupervised learning methods. Models tested on the pyrolysis data included linear, ridge, lasso, elastic-net, Gaussian process, random forest regression, and AutoSklearn, and the approach was continually refined to enhance predictive accuracy and model interpretability. Similar techniques were applied to the liquefaction data with an additional focus on feature engineering. Along with mesophase content, additional outputs of interest were the pitch yield, softening point, and QI content. Insights derived from these analyses are crucial in determining the factors influencing the quality and yield of coal-derived products. As the work progressed, the research evolved from foundational model comparisons to analyses of random forests, decision paths, and feature importance scores. A thorough market analysis was performed to examine the prospects of coal-based carbon fibers. The best opportunities for coal come from its lower and more stable price relative to petroleum, particularly for subbituminous coals, which is the primary advantage that a coal refinery may have over a petroleum refinery. Before a commercial CTP production facility can be modeled, however, several things need to be understood regarding the nature of the would-be coal refinery. These include the technology to be deployed, the size of facility, the volume(s) of co-product(s), and the waste and emissions profile of the plant. The volume of co-products and waste may be substantial and will require separate market analysis to ensure viability. In the near-term, the importance of coal tar pitch, in the form of carbon pitch, to the aluminum and steel industries is likely to overshadow the alternative use of this material as an input for carbon fiber. The importance of steel and aluminum in building materials, and the need for carbon materials in their manufacturing, will ensure that demand for these products remains for the long run. In addition, carbon fiber may also be the best substitute for steel and aluminum well into the future. While society will eventually be able to shift production of much of its electricity needs to renewables, it will not be able to shift away from fossil fuels for production of high-strength construction and vehicular materials. Demand for carbon fiber is expected to increase quickly, but the volume of carbon fiber and the amount of coal that would be needed to produce even a sizeable share of this market may still be relatively small compared to current coal production. Thus, other coal-based products like graphene, graphite, carbon foams, resins, and carbon-based building products will play important roles in sustaining coal production as coal-fired power generation continues to decline.

01 COAL, LIGNITE, AND PEAT↗

Atikokan Digital Twin: Machine learning in a biomass energy system

The Atikokan Generating Station, operated by Ontario Power Generation, has a 200 MW, biomass-fired tower boiler that operates on a dispatch schedule with a five-minute cycle. The boiler is generally operated in the range of 40–100 MW using two of five burner levels. In order to optimize boiler performance, we propose the implementation of a unique digital twin. Our digital twin abstraction couples Bayesian inference from science-based models and from observations (machine learning) with decision theory to predict operating-variable set points that optimize the physical asset (the boiler) in the presence of uncertainty (artificial intelligence). We focus this paper on the continuous Bayesian machine learning part of the Atikokan Digital Twin; we discuss decision theory in a companion paper. We identify and learn about 12 operational, model, and measured-output parameters and their uncertainties from high-fidelity, science-based simulations of the Atikokan boiler and from the observed measurements at the power plant. Since the goal of the Atikokan Digital Twin is to implement it online in real time, we require fast function evaluations for the quantities of interest extracted from the simulations in the Bayesian analysis. We use Gaussian process regression/interpolation to create accurate, robust surrogate models. We define the Bayesian priors and likelihood function and solve for the posterior distributions of the 12 parameters. Here we then propagate these distributions (i.e., parameters with uncertainty) into the predicted distributions of 790 quantities of interest to learn about the relative importance of various sources of error including experimental, model, and operating-parameter errors.

09 BIOMASS FUELS↗

Scalable and efficient algorithms for the propagation of uncertainty from data through inference to prediction for large-scale problems, with application to flow of the Antarctic ice sheet

The majority of research on efficient and scalable algorithms in computational science and engineering has focused on the forward problem : given parameter inputs, solve the governing equations to determine output quantities of interest. In contrast, in this paper, we consider the broader question: given a (large-scale) model containing uncertain parameters, (possibly) noisy observational data, and a prediction quantity of interest, how do we construct efficient and scalable algorithms to (1) infer the model parameters from the data (the deterministic inverse problem ), (2) quantify the uncertainty in the inferred parameters (the Bayesian inference problem), and (3) propagate the resulting uncertain parameters through the model to issue predictions with quantified uncertainties (the forward uncertainty propagation problem )? We present efficient and scalable algorithms for this end-to-end, data-to-prediction process under the Gaussian approximation and in the context of modeling the flow of the Antarctic ice sheet and its effect on loss of grounded ice to the ocean. The ice is modeled as a viscous, incompressible, creeping, shear-thinning fluid. The observational data come from satellite measurements of surface ice flow velocity, and the uncertain parameter field to be inferred is the basal sliding parameter, represented by a heterogeneous coefficient in a Robin boundary condition at the base of the ice sheet. The prediction quantity of interest is the present-day ice mass flux from the Antarctic continent to the ocean. We show that the work required for executing this data-to-prediction process—measured in number of forward (and adjoint) ice sheet model solves—is independent of the state dimension, parameter dimension, data dimension, and the number of processor cores. The key to achieving this dimension independence is to exploit the fact that, despite their large size, the observational data typically provide only sparse information on model parameters. This property can be exploited to construct a low rank approximation of the linearized parameter-to-observable map via randomized SVD methods and adjoint-based actions of Hessians of the data misfit functional.

54 ENVIRONMENTAL SCIENCES↗

A Fast, Two-dimensional Gaussian Process Method Based on Celerite: Applications to Transiting Exoplanet Discovery and Characterization

Gaussian processes (GPs) are commonly used as a model of stochastic variability in astrophysical time series. In particular, GPs are frequently employed to account for correlated stellar variability in planetary transit light curves. The efficient application of GPs to light curves containing thousands to tens of thousands of data points has been made possible by recent advances in GP methods, including the celerite method. Here we present an extension of the celerite method to two input dimensions where, typically, the second dimension is small. This method scales linearly with the total number of data points when the noise in each large dimension is proportional to the same celerite kernel and only the amplitude of the correlated noise varies in the second dimension. We demonstrate the application of this method to the problem of measuring precise transit parameters from multiwavelength light curves and show that it has the potential to improve transit parameters measurements by orders of magnitude. Applications of this method include transit spectroscopy and exomoon detection, as well a broader set of astronomical problems.

79 ASTRONOMY AND ASTROPHYSICS↗

Prediction of hydrogen adsorption in nanoporous materials from the energy distribution of adsorption sites

We present a fast and accurate, semi-analytical method for predicting hydrogen adsorption in nanoporous materials. For any temperature and pressure, the adsorbed amount is calculated as an integral over the energy density of adsorption sites (guest-host interactions) plus an average guest-guest term. The guest-host interaction energy is calculated using a classical force field with hydrogen modelled as a single-site probe. The guest-guest interaction energy is approximated using an average coordination number, which is regressed using Gaussian Process Regression (GPR). Local adsorption at each site is then modelled using a Langmuir isotherm, which when weighted with its probability density gives an accurate description of hydrogen adsorption. The method is tested on 933 metal-organic frameworks (MOFs) from the Computation-Ready Experimental (CoRE) MOF database at 77 K from 10 –5 to 100 bar, and the results are compared against GCMC predictions. To demonstrate the utility of the method, we calculated hydrogen adsorption isotherms for 12,914 existing MOF structures, at two different temperatures at a speed about 100 times that of GCMC simulations and analyzed the results. Here, we found 13 MOFs with predicted deliverable capacities exceeding the DOE target of 50 g/L for adsorption at 100 bar, 77 K and desorption at 5 bar, 160 K.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Constructing a Simulation Surrogate with Partially Observed Output

Gaussian process surrogates are a popular alternative to directly using computationally expensive simulation models. When the simulation output consists of many responses, dimension-reduction techniques are often employed to construct these surrogates. However, surrogate methods with dimension reduction generally rely on complete output training data. This article proposes a new Gaussian process surrogate method that permits the use of partially observed output while remaining computationally efficient. The new method involves the imputation of missing values and the adjustment of the covariance matrix used for Gaussian process inference. The resulting surrogate represents the available responses, disregards the missing responses, and provides meaningful uncertainty quantification. In conclusion, the proposed approach is shown to offer sharper inference than alternatives in a simulation study and a case study where an energy density functional model that frequently returns incomplete output is calibrated.

42 ENGINEERING↗

A BAYESIAN METHOD FOR THE ANALYSIS OF THE DUST EMISSION IN THE FAR-INFRARED AND SUBMILLIMETER

We present a method, based on Bayesian statistics, to fit the dust emission parameters in the far-infrared and submillimeter wavelengths. The method estimates the dust temperature and spectral emissivity index, plus their relationship, properly taking into account the statistical and systematic uncertainties. We test it on three sets of simulated sources detectable by the Herschel Space Observatory in the PACS and SPIRE spectral bands (70-500 {mu}m), spanning over a wide range of dust temperatures. The simulated observations are a one-component interstellar medium and two two-component sources, both warm (H II regions) and cold (cold clumps (CCs)). We first define a procedure to identify the better model, then we recover the parameters of the model and measure their physical correlations by means of a Markov Chain Monte Carlo algorithm adopting multi-variate Gaussian priors. In this process, we assess the reliability of the model recovery and of parameter estimation. We conclude that the model and parameters are properly recovered only under certain circumstances and that false models may be derived in some cases. We applied the method to a set of 91 starless CCs in an interarm region of the Galactic plane with low star formation activity, observed by Herschel in the Hi-GAL survey. Our results are consistent with a temperature-independent spectral index.

79 ASTRONOMY AND ASTROPHYSICS↗

Photometric redshifts for the S-PLUS Survey: Is machine learning up to the task?

The Southern Photometric Local Universe Survey (S-PLUS) is a novel project that aims to map the Southern Hemisphere using a twelve filter system, comprising five broad-band SDSS-like filters and seven narrow-band filters optimized for important stellar features in the local universe. In this paper we use the photometry and morphological information from the first S-PLUS data release (S-PLUS DR1) cross-matched to unWISE data and spectroscopic redshifts from Sloan Digital Sky Survey DR15. We explore three different machine learning methods (Gaussian Processes with GPz and two Deep Learning models made with TensorFlow) and compare them with the currently used template-fitting method in the S-PLUS DR1 to address whether machine learning methods can take advantage of the twelve filter system for photometric redshift prediction. Using tests for accuracy for both single-point estimates such as the calculation of the scatter, bias, and outlier fraction, and probability distribution functions (PDFs) such as the Probability Integral Transform (PIT), the Continuous Ranked Probability Score (CRPS) and the Odds distribution, we conclude that a deep-learning method using a combination of a Bayesian Neural Network and a Mixture Density Network offers the most accurate photometric redshifts for the current test sample. In conclusion, it achieves single-point photometric redshifts with scatter (σ NMAD ) of 0.023, normalized bias of -0.001, and outlier fraction of 0.64% for galaxies with r_auto magnitudes between 16 and 21.

79 ASTRONOMY AND ASTROPHYSICS↗

Investigation of Isobaric Mixing as a Mechanism for Boundary‐Layer Cloud Formation

This study investigates the potential role of isobaric mixing in the formation of marine boundary layer clouds. Cloud formation theory emphasizes uplift and adiabatic cooling, but recent observations support the existence of small clouds forming at various altitudes, even below the lifting condensation level. Isobaric mixing of air with different thermodynamic properties can generate localized supersaturation. A Gaussian mixing model is employed to simulate this process, considering the correlation between temperature and water vapor. Cloud droplet size distributions from aircraft measurements show a persistent and prominent mode of small droplets at 9 m, and the size of this mode compares favorably with predictions from the model. The results suggests that isobaric mixing plausibly contributes to the formation of clouds, particularly those observed at multiple altitudes with narrow droplet size distributions. This finding highlights the importance of considering isobaric mixing processes in understanding and modeling cloud formation.

54 ENVIRONMENTAL SCIENCES↗

Fission yield calculation using toy model based on Monte Carlo simulation

Toy model is a new approximation in predicting fission yield distribution. Toy model assumes nucleus as an elastic toy consist of marbles. The number of marbles represents the number of nucleons, A. This toy nucleus is able to imitate the real nucleus properties. In this research, the toy nucleons are only influenced by central force. A heavy toy nucleus induced by a toy nucleon will be split into two fragments. These two fission fragments are called fission yield. In this research, energy entanglement is neglected. Fission process in toy model is illustrated by two Gaussian curves intersecting each other. There are five Gaussian parameters used in this research. They are scission point of the two curves (R{sub c}), mean of left curve (μ{sub L}) and mean of right curve (μ{sub R}), deviation of left curve (σ{sub L}) and deviation of right curve (σ{sub R}). The fission yields distribution is analyses based on Monte Carlo simulation. The result shows that variation in σ or µ can significanly move the average frequency of asymmetry fission yields. This also varies the range of fission yields distribution probability. In addition, variation in iteration coefficient only change the frequency of fission yields. Monte Carlo simulation for fission yield calculation using toy model successfully indicates the same tendency with experiment results, where average of light fission yield is in the range of 90<A<100 and average of heavy fission yield is in about 135<A<145.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Direct nonparametric multimessenger constraints on the equation of state of cold dense nuclear matter

We utilize the now substantial amount of astrophysical observations of neutron stars (NSs), along with perturbative quantum chromodynamics (pQCD) calculations at high density, to directly constrain the NS equation of state (EOS). To this end, we construct nonparametric EOS priors by using Gaussian processes trained on 75 EOSs, which include models with either hadrons, hyperons, or quarks at high densities. We create a prior using the full EOS sample (model agnostic), and one prior for each EOS family to test model discrimination. We introduce a novel inference approach, which allows the simultaneous sampling of intrinsic and extrinsic parameters of binary NS mergers, as well as a nonparametric equation of state. We showcase this method in a Bayesian updating scheme by first performing a complete analysis of the binary NS merger event GW170817 with minimal assumptions, and sequentially adding information from x-ray and radio NS observations, along with pQCD calculations. Besides providing standard constraints, such as the pressure at twice nuclear saturation density 𝑝⁡(2⁢𝜌 sat ) = 4.3$^{+0.6}_{−0.6}$ × 10 34 dyne/cm 2 , at 95% confidence level, for the model agnostic prior, our methodology shows how the choice of EOS families used in conditioning changes the inferred astrophysical properties of the EOS, namely tidal deformability and maximum supported NS mass. We find hyperonic priors predicting higher tidal deformabilities for a 1.4⁢𝑀 ⊙ NS, and hadronic priors being preferred by the considered astrophysical data.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Descriptor Aided Bayesian Optimization for Many-Level Qualitative Variables With Materials Design Applications

Abstract Engineering design often involves qualitative and quantitative design variables, which requires systematic methods for the exploration of these mixed-variable design spaces. Expensive simulation techniques, such as those required to evaluate optimization objectives in materials design applications, constitute the main portion of the cost of the design process and underline the need for efficient search strategies—Bayesian optimization (BO) being one of the most widely adopted. Although recent developments in mixed-variable Bayesian optimization have shown promise, the effects of dimensionality of qualitative variables have not been well studied. High-dimensional qualitative variables, i.e., with many levels, impose a large design cost as they typically require a larger dataset to quantify the effect of each level on the optimization objective. We address this challenge by leveraging domain knowledge about underlying physical descriptors, which embody the physics of the underlying physical phenomena, to infer the effect of unobserved levels that have not been sampled yet. We show that physical descriptors can be intuitively embedded into the latent variable Gaussian process approach—a mixed-variable GP modeling technique—and used to selectively explore levels of qualitative variables in the Bayesian optimization framework. This physics-informed approach is particularly useful when one or more qualitative variables are high dimensional (many-level) and the modeling dataset is small, containing observations for only a subset of levels. Through a combination of mathematical test functions and materials design applications, our method is shown to be robust to certain types of incomplete domain knowledge and significantly reduces the design cost for problems with high-dimensional qualitative variables.

Engineering↗

Model independent comparison of supernova and strong lensing cosmography: Implications for the Hubble constant tension

We use supernovae measurements, calibrated by the local determination of the Hubble constant $H_0$ by SH0ES, to interpolate the distance-redshift relation using Gaussian process regression. We then predict, independent of the cosmological model, the distances that are measured with strong lensing time delays to test their mutual agreement. In this work we find excellent agreement between these predictions and the measurements. The agreement holds when we consider only the redshift dependence of the distance-redshift relation, independent of the value of $H_0$. Our results disfavor the possibility that lens mass modeling contributes a 10% bias or uncertainty in the strong lensing analysis, as suggested recently in the literature. In general our analysis strengthens the case that residual systematic errors in both measurements are below the level of the current discrepancy with the CMB determination of $H_0$, and supports the possibility of new physical phenomena on cosmological scales. With additional data our methodology can provide more stringent tests of unaccounted for systematics in the determinations of the distance-redshift relation in the late universe.

79 ASTRONOMY AND ASTROPHYSICS↗

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↗