Search NASA⌕ Search

SEARCH · Search NASA

Results for “learning framework”

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 703 records · Page 39
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Intelligent Experiments Through Real-time AI: Fast Data Processing and Autonomous Detector Control for sPHENIX and Future EIC Detectors (Final Report)

The overall vision of this project was to integrate real-time artificial intelligence (AI) directly into the data acquisition and detector-control systems of nuclear physics experiments, including both fast online event selection and an autonomous detector-control feedback loop. The work carried out under the award focused on the fast online event-selection half of that vision: the efficient recording of low-momentum heavy-flavor (HF) hadron decays in proton-proton collisions at the sPHENIX experiment at the Relativistic Heavy Ion Collider (RHIC)—an observable that requires fast tracking and topological trigger selection not previously demonstrated at RHIC, and that is essential for QCD studies at future facilities such as the Electron-Ion Collider (EIC). The autonomous detector-control (GPU-based feedback) component named in the project title remained a design concept and was not implemented under this award. The Massachusetts Institute of Technology (MIT) group led the offline simulation and data processing needed to train the machine-learning (ML) models, the translation of trained models to Field-Programmable Gate Array (FPGA) firmware using the hls4ml framework, and the physics validation of heavy-flavor reconstruction. Over the award period, the team developed and hardware-tested the principal components of an AI-based heavy-flavor trigger on simulated and recorded sPHENIX tracker data: a software Bipartite Graph Attention Network (BiGAT) trigger model reaching > 95% signal efficiency at 99% background rejection; an FPGA-native hit clusterizer matching the offline clustering; smaller networks synthesized to FPGA within the required sub-10 µs latency; and an assembled decoder–clusterizer–inference firmware chain exercised on the FELIX readout board. A complete, fully integrated hardware demonstrator was not finished within the award period. This report documents the project goals, the MIT group’s contributions, the technical accomplishments, and the outlook toward applications at the future EIC ePIC detector.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

Idiomatic Correctness-Checking via Julienne in Fortran 2023

This paper presents a unified approach to unit testing and runtime assertion checking using Fortran 2023. The paper describes the support for our approach in the Julienne framework. Julienne leverages recent Fortran standards to implement object-oriented design patterns, support testing parallel programs, and implement functional programming patterns in order to craft idioms inspired by natural-language expressions. The presented idioms employ novel operators to write expressions that evaluate to a test-diagnosis object encapsulating two components: (1) the test outcome or assertion outcome and (2) an automatically generated diagnostic string. Two other novel aspects of the approach include (1) the ability to enforce assertions inside pure procedures and (2) the ability to output rich diagnostic information inside pure procedures during error termination when assertions fail. The latter capability mitigates against a reason that Fortran programmers commonly cite for not writing pure procedures: difficulty obtaining useful program output inside pure procedures when debugging code. This paper demonstrates how the adoption of the proposed idioms leads naturally to a unifying theme across two otherwise disparate technologies: unit testing and runtime assertion checking. Finally, this paper describes the usage of the Julienne testing framework for writing unit tests and assertions in the Matcha high-performance computing application and the Fiats deep learning library.

Rouson, Damian↗
DOE OSTI2025

Insights into distorted lamellar phases with small-angle scattering and machine learning

Lamellar phases are essential in various soft matter systems, with topological defects significantly influencing their mechanical properties. In this report, we present a machine-learning approach for quantitatively analyzing the structure and dynamics of distorted lamellar phases using scattering techniques. By leveraging the mathematical framework of Kolmogorov–Arnold networks, we demonstrate that the conformations of these distorted phases – expressed as superpositions of complex waves – can be reconstructed from small-angle scattering intensities. Through the contour analysis of wave field phase singularities, we obtain the statistics of the spatial distribution of topological defects. Furthermore, we establish that the temporal evolution of these defects can be derived from the time-dependent traveling wave field, informed by the dispersion relation of spectral components. This method opens new avenues for investigating the dynamics of distorted lamellar phases using various dynamic scattering techniques such as neutron spin echo and X-ray photon correlation spectroscopy. These findings enhance our microscopic understanding of how defects influence the physical properties of lamellar materials, with implications for both equilibrium and non-equilibrium states in general lamellar systems.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Generalization error guaranteed auto-encoder-based nonlinear model reduction for operator learning

Many physical processes in science and engineering are naturally represented by operators between infinite-dimensional function spaces. The problem of operator learning, in this context, seeks to extract these physical processes from empirical data, which is challenging due to the infinite or high dimensionality of data. An integral component in addressing this challenge is model reduction, which reduces both the data dimensionality and problem size. In this paper, we utilize low-dimensional nonlinear structures in model reduction by investigating Auto-Encoder-based Neural Network (AENet). AENet first learns the latent variables of the input data and then learns the transformation from these latent variables to corresponding output data. Our numerical experiments validate the ability of AENet to accurately learn the solution operator of nonlinear partial differential equations. Furthermore, we establish a mathematical and statistical estimation theory that analyzes the generalization error of AENet. Finally, our theoretical framework shows that the sample complexity of training AENet is intricately tied to the intrinsic dimension of the modeled process, while also demonstrating the robustness of AENet to noise.

Auto-encoder↗
DOE OSTI2026

Lessons Learned from Open-Source Software Training Toward Expanding the Fusion Workforce

Within the United States (U.S.), the Fusion Innovation Research Engine (FIRE) Collaboratives seek to accelerate fusion technology development through wide-ranging community-driven research activities that bring together industry, laboratories, research institutes, and academia. Increasing the maturity of key components and systems for future fusion power plants (FPPs) toward commercialization, will, by necessity, increase the need for knowledgeable engineers, designers, and researchers in industry capable of integrating and further improving these technologies within industry FPP concepts. Further, various modeling and simulation packages support these programs and are under-development within them to drive design iteration and the development of FPP digital twins. To derive the greatest benefit from model output and capabilities, training these same specialists will be vital. Thus, the development of effective and accessible software onboarding and professional development opportunities focused on fusion will be key to keeping the pace of growth high in the coming decade and beyond. A similar need exists within the advanced fission reactor community, driven by an ever-growing need for carbon-free baseload power for industrial and data center applications. Within the U.S. and around the world, several open-source packages and frameworks exist to support this endeavor, and two we will highlight here are the Multiphysics Object Oriented Simulation Environment (MOOSE) framework and OpenMC. These packages, notably, are also being utilized in the FIRE Collaboratives program. In this presentation, we will highlight the lessons learned from over 30 years of combined software development and training experience, focused on developing strong technical software foundations within the nuclear workforce, from students to professional engineers & scientists. We will connect this experience to present and emerging needs in the fusion energy community, and, finally, will outline possible paths forward to develop a large, robust, global fusion workforce.

70 - PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Developing a digital twin framework for remotely monitoring nuclear reactor facilities

A digital twin must seek to represent all applicable functional components of the system of interest. Different expertise is required for understanding the physical system being modeled than the skills needed for transforming those models into a functional digital twin through physics modeling, machine learning analysis, and visualization. The diversity of knowledge requires a multi-disciplinary team to ensure all system details are captured. Team members also need a method to verify that the data they generate within their domain can be effectively communicated to professionals in other fields. To address this challenge, this work provides an approach for developing a digital twin framework to remotely monitoring nuclear facilities. Through this, general knowledge of the framework is presented along with two examples to solidify the process. The AGN-201 digital twin and microreactor digital twins provide varying levels of complexity in a potential nuclear facility, where common threads are identified and lessons learned are provided. The goal of this research is to aid future researchers by providing a formula for a successful digital twin and in turn reducing the development time of nuclear system digital twins, specifically for remote monitoring.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

Scalability analysis of heavy-duty gas turbines using data-driven machine learning

With the increasing integration of variable renewable energy sources into power systems, the role of flexible power generation technologies like gas turbines (GT) in rapid grid balancing remains crucial. This sustained importance underscores the need for scaled and precise modeling of GT to ensure effective integration within evolving energy frameworks. While physics-driven GT models integrate thermodynamics, fluid dynamics, and combustion principles, they often rely on approximate mathematical representations to accommodate scaling that may not capture the actual complex dynamics for GTs and inertial effects associated to GTs with different ratings. In this study, a data-driven model is proposed using machine learning (ML) techniques to conduct GT scalability analysis and performance evaluation with high accuracy. The ML model, trained on data from various operating conditions and performance parameters, aims to uncover intricate relationships and patterns, resembling GT characteristics at different scales (ratings). The model is developed to capture complex system interaction and to adapt to changing operational scenarios at different capacities, providing valuable insights of power system dynamics. In this study, the real-time digital simulator platform was employed to generate training data for the ML model and assess its dynamic characteristics. The ultimate objective was to develop a detailed modeling framework based on governing equations and data-driven ML capable of predicting key performance indicators, in thermal systems such as GTs, including power output, speed, fuel consumption, and exhaust temperature under diverse operating conditions at different scales. The developed ML framework demonstrated high accuracy, with mean relative errors for GT power prediction, reference speed, exhaust temperature, and compressor pressure ratio (CPR) parameters consistently below 0.1% across typical load fluctuation scenarios. Maximum deviations were limited to approximately 0.5 K for exhaust temperature and 0.009 for CPR, underscoring the model’s ability to replicating dynamic GT behavior with high precision. The adaptability of the ML model enables its application across diverse operational conditions and its extension to other thermal systems. By leveraging advanced ML techniques, this study presents a robust and scalable modeling framework that enhances GT simulation precision, facilitating improved integration into evolving power systems.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

A general Bayesian algorithm for the autonomous alignment of beamlines

Autonomous methods to align beamlines can decrease the amount of time spent on diagnostics, and also uncover better global optima leading to better beam quality. The alignment of these beamlines is a high-dimensional expensive-to-sample optimization problem involving the simultaneous treatment of many optical elements with correlated and nonlinear dynamics. Bayesian optimization is a strategy of efficient global optimization that has proved successful in similar regimes in a wide variety of beamline alignment applications, though it has typically been implemented for particular beamlines and optimization tasks. In this paper, we present a basic formulation of Bayesian inference and Gaussian process models as they relate to multi-objective Bayesian optimization, as well as the practical challenges presented by beamline alignment. We show that the same general implementation of Bayesian optimization with special consideration for beamline alignment can quickly learn the dynamics of particular beamlines in an online fashion through hyperparameter fitting with no prior information. We present the implementation of a concise software framework for beamline alignment and test it on four different optimization problems for experiments on X-ray beamlines at the National Synchrotron Light Source II and the Advanced Light Source, and an electron beam at the Accelerator Test Facility, along with benchmarking on a simulated digital twin. We discuss new applications of the framework, and the potential for a unified approach to beamline alignment at synchrotron facilities.

47 OTHER INSTRUMENTATION↗
DOE OSTI2026

Explainable Graph Learning for Particle Accelerator Operations

Particle accelerators are vital tools in physics, medicine, and industry, requiring precise tuning to ensure optimal beam performance. However, real-world deviations from idealized simulations make beam tuning a time-consuming and error-prone process. In this work, we propose an explanation-driven framework for providing actionable insight into beamline operations, with a focus on the injector beamline at the Continuous Electron Beam Accelerator Facility (CEBAF). We represent beamline configurations as heterogeneous graphs, where setting nodes represent elements that human operators can actively adjust during beam tuning, and reading nodes passively provide diagnostic feedback. To identify the most influential setting nodes responsible for differences between any two beamline configurations, our approach first predicts the resulting changes in reading nodes caused by variations in settings, and then learns importance scores that capture the joint influence of multiple setting nodes. Experimental results on real-world CEBAF injector data demonstrate the framework’s ability to generate interpretable insights that can assist human operators in beamline tuning and reduce operational overhead.

Wang, Song [Univ. of Virginia, Charlottesville, VA↗
DOE OSTI2024

Generative deep-learning reveals collective variables of Fermionic systems

Complex processes of fermionic systems ranging from protein folding to nuclear fission often follow a low-dimensional reaction path parametrized in terms of a few collective variables. In nuclear theory, variables related to the shape of the nuclear density in a mean-field picture are key to describing the large amplitude collective motion of the neutrons and protons. Exploring the adiabatic energy landscape spanned by these degrees of freedom reveals the possible reaction channels while simulating the dynamics in this reduced space yields their respective probabilities. Unfortunately, this theoretical framework breaks down whenever the systems encounters a quantum phase transition with respect to the collective variables. Here, in this study, we introduce a novel generative deep-learning algorithm designed to build reaction paths that ensure that the many-fermion wave function stays differentiable with respect to the collective variables. This approach is applicable to any fermionic system described by a coherent state. We use the case of potential energy curves in the 16 O nucleus within the Hartree-Fock theory to illustrate its main features.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2026

EC-Bench: A Benchmark for Enzyme Commission Number Prediction

Enzymes are proteins that catalyze specific biochemical reactions in cells. Enzyme Commission (EC) numbers are used to annotate enzymes in a four-level hierarchy that classifies enzymes based on the specific chemical reactions they catalyze. Accurate EC number prediction is essential for understanding enzyme functions. Despite the availability of numerous methods for predicting EC numbers from protein sequences, there is no unified framework for evaluating and studying such methods systematically. This gap limits the ability of the community to identify the most effective approaches for enzyme annotation. We introduce EC-Bench, a benchmark for EC number prediction, consisting of 1) an initial representative set of existing methods (including homology-based, deep learning, contrastive learning, and language model methods), 2) existing and novel accuracy and efficiency performance metrics, and 3) selected datasets to allow for comprehensive comparative study. EC-Bench is open-source and provides a framework for researchers to not only compare among existing methods objectively under uniform conditions, but also to introduce and effectively evaluate performance of new methods in a comparative framework. To demonstrate the utility of EC-Bench, we perform extensive experimentation to compare the existing EC number prediction methods and establish their advantages and disadvantages in a variety of prediction tasks, namely “exact EC number prediction”, “EC number completion” and (partial or additional) “EC number recommendation”. We find wide variation in the performance of different methods, but also subtle but potentially useful differences in the performance of different methods across tasks and for different parts of the EC hierarchy.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

ADEPT: A Pedagogical Framework for Integrating Agentic AI with Deterministic Scientific Workflows

The integration of Large Language Models (LLMs) into scientific research promises to accelerate discovery, yet a significant gap remains between the dynamic reasoning of Artificial Intelligence (AI) agents and the static, deterministic nature of canonical scientific workflows. This paper introduces ADEPT (Agentic Discovery and Exploration Platform for Tools), a reference architecture and pedagogical framework explicitly designed to bridge this gap. ADEPT's primary mission is to provide a transparent, "glass-box" environment where researchers and engineers can learn to effectively wrap established scientific software (e.g., BLAST, Nextflow pipelines) and compose it into reliable, agent-driven workflows. We describe its modular, multi-server architecture, which leverages the Model Context Protocol (MCP) for tool serving, LangGraph for robust agentic orchestration, and a secure nsjail-based sandbox for safe code execution. By prioritizing architectural clarity, safety, and modularity, ADEPT serves as an extensible blueprint for building trustworthy AI-augmented systems and fosters the collaborative development necessary to responsibly employ agentic AI for science. We provide practical examples of how to adapt and extend this framework, highlighting its utility in workforce development and AI-readiness capabilities across research and development projects.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

On learning what to learn: Heterogeneous observations of dynamics and establishing possibly causal relations among them

Abstract Before we attempt to (approximately) learn a function between two sets of observables of a physical process, we must first decide what the inputs and outputs of the desired function are going to be. Here we demonstrate two distinct, data-driven ways of first deciding “the right quantities” to relate through such a function, and then proceeding to learn it. This is accomplished by first processing simultaneous heterogeneous data streams (ensembles of time series) from observations of a physical system: records of multiple observation processes of the system. We determine (i) what subsets of observables are common between the observation processes (and therefore observable from each other, relatable through a function); and (ii) what information is unrelated to these common observables, therefore particular to each observation process, and not contributing to the desired function. Any data-driven technique can subsequently be used to learn the input–output relation—from k-nearest neighbors and Geometric Harmonics to Gaussian Processes and Neural Networks. Two particular “twists” of the approach are discussed. The first has to do with the identifiability of particular quantities of interest from the measurements. We now construct mappings from a single set of observations from one process to entire level sets of measurements of the second process, consistent with this single set. The second attempts to relate our framework to a form of causality: if one of the observation processes measures “now,” while the second observation process measures “in the future,” the function to be learned among what is common across observation processes constitutes a dynamical model for the system evolution.

Sroczynski, David W.↗
DOE OSTI2024

Fourier-MIONet: Fourier-enhanced multiple-input neural operators for multiphase modeling of geological carbon sequestration

Geologic carbon sequestration (GCS) is a safety-critical technology that aims to reduce the amount of carbon dioxide in the atmosphere, which also places high demands on reliability. Multiphase flow in porous media is essential to understand CO 2 migration and pressure fields in the subsurface associated with GCS. However, numerical simulation for such problems in 4D is computationally challenging and expensive, due to the multiphysics and multiscale nature of the highly nonlinear governing partial differential equations (PDEs). It prevents us from considering multiple subsurface scenarios and conducting real-time optimization. Here, we develop a Fourier-enhanced multiple-input neural operator (Fourier-MIONet) to learn the solution operator of the problem of multiphase flow in porous media. Fourier-MIONet utilizes the recently developed framework of the multiple-input deep neural operators (MIONet) and incorporates the Fourier neural operator (FNO) in the network architecture. Once Fourier-MIONet is trained, it can predict the evolution of saturation and pressure of the multiphase flow under various reservoir conditions, such as permeability and porosity heterogeneity, anisotropy, injection configurations, and multiphase flow properties. Compared to the enhanced FNO (U-FNO), the proposed Fourier-MIONet has 90% fewer unknown parameters, and it can be trained in significantly less time (about 3.5 times faster) with much lower CPU memory (<15%) and GPU memory (<35%) requirements, to achieve similar prediction accuracy. In addition to the lower computational cost, Fourier-MIONet can be trained with only 6 snapshots of time to predict the PDE solutions for 30 years. Furthermore, we observed that Fourier-MIONet can maintain good accuracy when predicting out-of-distribution (OOD) data. The excellent generalizability of Fourier-MIONet is enabled by its adherence to the physical principle that the solution to a PDE is continuous over time. Furthermore, the developed Fourier-MIONet makes it possible to solve the long-time evolution of geological carbon sequestration in a large-scale three-dimensional space accurately and efficiently.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Data mining and computational screening of Rashba-Dresselhaus splitting and optoelectronic properties in two-dimensional perovskite materials

Recent developments highlighting the promise of two-dimensional perovskites have vastly increased the compositional search space in the perovskite family. This presents a great opportunity for the realization of highly performant devices and practical challenges associated with the identification of candidate materials. High-fidelity computational screening offers great value in this regard. In this study, we carry out a multiscale computational workflow, generating a dataset of two-dimensional perovskites in the Dion-Jacobson and Ruddlesden-Popper phases. Our dataset comprises ten B-site cations, four halogens, and over 20 organic cations across over 2000 materials. We compute electronic properties, thermoelectric performance, and numerous geometric characteristics. Furthermore, we introduce a framework for the high-throughput computation of Rashba-Dresselhaus splitting. Finally, we use this dataset to train machine learning models for the accurate prediction of band gaps, candidate Rashba-Dresselhaus materials, and partial charges. The work presented herein can aid future investigations of two-dimensional perovskites with targeted applications in mind.

14 SOLAR ENERGY↗
DOE OSTI2025

Statistical relationships across epigenomes using large-scale hierarchical clustering

Recent advances in genomics and sequencing platforms have revolutionized our ability to create immense data sets, particularly for studying epigenetic regulation of gene expression. However, the avalanche of epigenomic data is difficult to parse for biological interpretation given nonlinear complex patterns and relationships. This attractive challenge in epigenomic data lends itself to machine learning for discerning infectivity and susceptibility. In this study, we explore over 3000 epigenomes of uninfected individuals and provide a framework to characterize the relationships among epigenetic modifiers, their modifiers, genetic loci, and specific immune cell types across all chromosomes using hierarchical clustering. Hierarchical clustering of epigenomic data revealed consistent epigenetic patterns across chromosomes, demonstrating that variation due to epigenetic modifiers is greater than variation between cell types. Gene Ontology and KEGG pathway analyses indicated significant enrichment of genes involved in chromatin remodeling, mRNA splicing, immune responses, and the regulation of microRNAs and snoRNAs. Epigenetic modifiers frequently formed biologically relevant clusters, including the cohesin complex, RNA Polymerase II transcription factors, and PRC2 complex members. These clustering behaviors remained consistent across all chromosomes, supported by entropy analysis and high Adjusted Rand Index scores, indicating robust cross-chromosomal similarity. Co-occurrence analysis further revealed specific sets of modifiers that consistently appeared together within clusters, reflecting shared biological functions and interactions. Validation using another dataset confirmed the reproducibility of these clustering patterns and modifier co-occurrence relationships, underscoring the reliability and generalizability of the methodology.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Hybrid learning techniques for scientific data reduction with performance guarantees

The research initiatives supported by the U.S. Department of Energy (DOE) Grant DE-SC0022265 are fundamentally aimed at pioneering advanced machine learning (ML) techniques for scientific data compression within high-performance computing (HPC) environments. This comprehensive body of work addresses the critical challenge posed by the exponential growth of data generated by scientific simulations in domains such as fusion energy, climate modeling, and computational fluid dynamics (CFD). A core objective is to develop compression algorithms that achieve substantial data reduction—often by orders of magnitude—while rigorously ensuring the fidelity of both the primary data (PD) and scientifically crucial derived quantities of interest (QoI). The methodologies deployed under this grant integrate sophisticated deep learning architectures, prominently featuring autoencoders, advanced generative models like conditional diffusion, and hybrid learning techniques. Key innovations include the development of Guaranteed Autoencoders (GAE) and the Guaranteed Conditional Diffusion with Tensor Correction (GCDTC) framework, which provide explicit, instance-level error bounds on reconstructed data. Furthermore, specialized strategies such as nonlinear constraint satisfaction are employed to preserve the integrity of QoI, a vital requirement for the trustworthiness of downstream scientific analyses. This research also focuses on the design and implementation of scalable, GPU-accelerated software pipelines that seamlessly integrate into existing HPC workflows, ensuring both computational efficiency and practical applicability. The CAESAR framework, for example, unifies foundation and generative models to create an adaptive and efficient compression solution for spatio-temporal scientific data. Collectively, these efforts represent a significant advancement in mitigating the scientific data deluge, enabling more effective data management, accelerated scientific discovery, and optimized utilization of HPC resources.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Knowledge Oriented Graph Unified Transformer (KOGUT) v0.1

KOGUT — Knowledge Oriented Graph Unified Transformer KOGUT implements the Relational Graph Transformer (RelGT) architecture for knowledge graph link prediction in biological domains, with a primary focus on microbial growth media prediction. While the original RelGT (arXiv:2505.10960) targets relational tables, time series, and multi-table databases, KOGUT adapts this architecture for heterogeneous biological knowledge graphs, providing first-in-class AI predictive models for microbial cultivation. Key Adaptations Beyond Original RelGT: - Knowledge Graph Focus: Applied to biological KGs with semantic node types (taxa, chemicals, media, phenotypes, environments) versus generic relational database tables, trained on the KG-Microbe knowledge graph (1.3M entities, 2.9M edges, 24 relation types). - Multimodal Node Encoding: Integrates node labels, categories, descriptions, and synonyms from KG metadata through learned embedding layers—adapting relational column features to graph node attributes with textual semantics. - Extended K-Hop Subgraph Strategy: Optimized neighborhood sampling (3-hop default, configurable up to 200 nodes) tuned for sparse biological networks, building on the original local-global attention framework with biological relation preservation. - Biolink Predicate Preservation: Type-specific transformations for 24 biological edge semantics (occurs_in, consumes, produces, has_phenotype, subclass_of) beyond standard relational foreign keys, enabling multi-relation link prediction. - Inductive Learning Support: Enables zero-shot predictions for novel taxa through feature-based embeddings (temperature, oxygen requirements, gram stain, cell shape), extending the original transductive relational benchmark scope to uncultured microorganisms. CheapSOTA Performance Optimizations (This Distribution): - VQ-EMA Centroid Attention: Vector quantization with exponential moving average for improved global context modeling (+5-10% MRR improvement). - HDF5 Precomputed Data Loading: One-time preprocessing of k-hop subgraphs to eliminate redundant graph traversals (2-5× training speedup). - Distributed Data Parallel Training: Multi-GPU support for scaling to larger knowledge graphs (tested on 4× NVIDIA A100 GPUs at NERSC Perlmutter). - Mixed Precision Training: Automatic mixed precision (AMP) for memory efficiency and faster training. Advantages Over Standard Knowledge Graph Embedding Models: Combines RelGT's proven multi-element tokenization (features, type, hop, structure) with graph-native biological representations, enabling interpretable link prediction across heterogeneous entities that standard embedding models (TransE, RotatE, ComplEx) and table-based transformers cannot directly model. Achieves near-perfect performance on microbial growth media prediction (MRR: 0.9966, Precision@1: 0.9932, Hit@10: 1.0000) while maintaining explainability through attention-based reasoning over biological pathways. Training Data: - KG-Microbe merged knowledge graph: 1,379,337 nodes, 2,960,472 edges - 24 biological relation types including taxonomic hierarchies, metabolic interactions, phenotype associations, and environmental relationships - Primary prediction task: Growth media suitability for microbial taxa (biolink:occurs_in, 50K edges) - Multi-relation capability: Predicts links for any of the 24 relation types, including chemical consumption/production, phenotype associations, and taxonomic classification Citation: Original RelGT Architecture: Dwivedi et al., "Relational Graph Transformer", arXiv:2505.10960, 2025 KOGUT Implementation: Knowledge Oriented Graph Unified Transformer for Microbial Growth Media Prediction Developed at Lawrence Berkeley National Laboratory (LBNL) Trained on NERSC Perlmutter supercomputer

Joachimiak, Marcin [Lawrence Berkeley National Lab↗