Interpretable A-posteriori error indication for graph neural network surrogate models
Not Available
Barwey, Shivam (ORCID:0000000217171805)↗
SEARCH · Search NASA
Results for “graph model”
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.
Graph Convolutional Neural Network Modeling of Vacancy Formation for Materials Discovery in Solar Thermochemical Water Splitting.
Abstract not provided.
Witman, Matthew↗
Graph Neural Networks for Surrogate Modeling of Offshore Floating Platforms
Floating offshore wind turbines (FOWTs) present an significant opportunity to increase renewable energy generation. However, significant challenges remain before FOWTs can be widely commercialized and deployed. In particular, hydrodynamic loading on the platforms can stress the overall structure, damage the mooring systems, and impact power generation. Studying these loads is difficult and often relies on computationally expensive models or experiments. In this work, we explore the use of graph neural networks (GNNs) to construct flexible, data-driven surrogates for hydrodynamic loads on platforms. We leverage the natural graph-like structure of offshore wind platform designs to enable the GNN model to learn to approximate the loads for different wave conditions and structural designs. We demonstrate potential uses for the surrogate by performing parameter sweeps and ridge analysis on the trained model to identify the impacts of different wave and structural features on the loads.
floating offshore wind turbines↗
On the degeneracy of spin ice graphs, and its estimate via the Bethe permanent
The concept of spin ice can be extended to a general graph. We study the degeneracy of spin ice graph on arbitrary interaction structures via graph theory. We map spin ice graphs to the Ising model on a graph and clarify whether the inverse mapping is possible via a modified Krausz construction. From the gauge freedom of frustrated Ising systems, we derive exact, general results about frustration and degeneracy. We demonstrate for the first time that every spin ice graph, with the exception of the one-dimensional Ising model, is degenerate. We then study how degeneracy scales in size, using the mapping between Eulerian trails and spin ice manifolds, and a permanental identity for the number of Eulerian orientations. Furthermore, we show that the Bethe permanent technique provides both an estimate and a lower bound to the frustration of spin ices on arbitrary graphs of even degree. While such a technique can also be used to obtain an upper bound, we find that in all finite degree examples we studied, another upper bound based on Schrijver inequality is tighter.
97 MATHEMATICS AND COMPUTING↗
pnnl/grid_prediction
Two datadriven predictive approaches, namely, {\em Koopman Operator Theoretic (KOT)-based} model and {\em Graph Neural Network (GNN)-based model}, to enable effective power system state predictions. The KOT-based approaches (Robust DMD, deepDMD) capture the power system evolution as a linear dynamical system on an abstract space. The GNNs model the spatio-temporal correlations using graph convolutional network and are called Spatio-Temporal Graph Convolutional Network (STGCN). These predictive models are trained, tested and compared rigorously based on their predictions of frequencies in the IEEE 68 bus system when subjected to a disturbance. GridSTAGE framework developed at Pacific Northwest National Laboratory is leveraged to generate multiple datasets (in the form of PMU measurements) for training and testing by strategically creating load changes across the spatial locations of the network
Nandanoor, Sai Pushpak↗
Knowledge Graph Embedding using Large Language Models for COVID-19
CoVID-19 (cord19) open data source for training knowledge graph with LLM
Cord19↗
3D-equivariant graph neural networks for protein model quality assessment
Quality assessment (QA) of predicted protein tertiary structure models plays an important role in ranking and using them. With the recent development of deep learning end-to-end protein structure prediction techniques for generating highly confident tertiary structures for most proteins, it is important to explore corresponding QA strategies to evaluate and select the structural models predicted by them since these models have better quality and different properties than the models predicted by traditional tertiary structure prediction methods. We develop EnQA, a novel graph-based 3D-equivariant neural network method that is equivariant to rotation and translation of 3D objects to estimate the accuracy of protein structural models by leveraging the structural features acquired from the state-of-the-art tertiary structure prediction method—AlphaFold2. We train and test the method on both traditional model datasets (e.g. the datasets of the Critical Assessment of Techniques for Protein Structure Prediction) and a new dataset of high-quality structural models predicted only by AlphaFold2 for the proteins whose experimental structures were released recently. Our approach achieves state-of-the-art performance on protein structural models predicted by both traditional protein structure prediction methods and the latest end-to-end deep learning method—AlphaFold2. It performs even better than the model QA scores provided by AlphaFold2 itself. The results illustrate that the 3D-equivariant graph neural network is a promising approach to the evaluation of protein structural models. Integrating AlphaFold2 features with other complementary sequence and structural features is important for improving protein model QA.
59 BASIC BIOLOGICAL SCIENCES↗
GSplit: Scaling Graph Neural Network Training on Large Graphs via Split-Parallelism
Graph neural networks (GNNs), an emerging class of machine learning models for graphs, have gained popularity for their superior performance in various graph analytical tasks. Mini-batch training is commonly used to train GNNs on large graphs, and data parallelism is the standard approach to scale mini-batch training across multiple GPUs. Data parallel approaches contain redundant work as subgraphs sampled by different GPUs contain significant overlap. To address this issue, we introduce a hybrid parallel mini-batch training paradigm called Split parallelism. Split parallelism avoids redundant work by splitting the sampling, loading, and training of each mini-batch across multiple GPUs. Split parallelism, however, introduces communication overheads that can be more than the savings from removing redundant work. We further present a lightweight partitioning algorithm that probabilistically minimizes these overheads. We implement spllit parllelism in GSplit and show that it outperforms state-of-the-art mini-batch training systems like DGL, Quiver, and P3.
Lim, Seung-Hwan [ORNL] (ORCID:0000000194616866)↗
DG2DAG: Learning Directed Acyclic Graphs from Functional Priors
Physics-based systems-of-systems models are computationally expensive. Reduced graphical models can decrease computational complexity, but may not proffer an end-to-end model from upstream inputs to downstream outputs. We consequently are interested in reducing models on directed graphs to models on a directed acyclic subgraph such that preserves accurate reconstruction of nodes. The consequence is a model with a topological ordering, providing a one-way flow of computation, and a causal interpr
Voronin, Alexey [Sandia National Laboratories (SNL↗
Anticipating Technical Expertise and Capability Evolution in Research Communities Using Dynamic Graph Transformers
The ability to anticipate global technical expertise and capability evolution trends is essential for national and global security, especially in safety-critical domains such as nuclear nonproliferation (NN) and rapidly emerging fields like artificial intelligence (AI). Here, in this work, we extend traditional statistical relational learning approaches (e.g., link prediction in collaboration networks) and formulate a problem of anticipating technical expertise and capability evolution using dynamic heterogeneous graph representations. We develop novel capabilities to forecast collaboration patterns, authorship behavior, and technical capability evolution at different granularities (e.g., scientist and institution levels) in two distinct research fields. We implement a dynamic graph transformer (DGT) neural architecture, which pushes the state-of-the-art graph neural network models by: 1) forecasting heterogeneous (rather than homogeneous) nodes and edges; and 2) relying on both discrete- and continuous-time inputs. We demonstrate that our DGT models predict collaboration, partnership, and expertise patterns with 0.26, 0.73, and 0.53 mean reciprocal rank values for AI and 0.48, 0.93, and 0.22 for NN domains. DGT model performance exceeds the best-performing static graph baseline models by 30%–80% across AI and NN domains. Our findings demonstrate that DGT models boost inductive task performance when previously unseen nodes appear in the test data for the domains with emerging collaboration patterns (e.g., AI). Specifically, models accurately predict which established scientists will collaborate with early career scientists and vice versa in the AI domain.
97 MATHEMATICS AND COMPUTING↗
Applying machine learning and quantum chemistry to predict the glass transition temperatures of polymers
Glass transition temperature (T g ) is important for understanding the physical and mechanical properties of a polymer material because it relates to the thermal energy required to transition between a hard glassy state and a soft rubbery one. Over the years, various models have been developed for predicting this thermal property from molecular structure to aid in designing novel polymers in selected classes. This work builds on those efforts by utilizing both machine learning (ML) and quantum chemistry (QC) techniques to develop models that can predict T g values from the molecular structure under different data availability scenarios and for a wide variety of polymer types. For the ML model, a graph convolutional network (GCN) was used to map topological polymer features; this model was trained against a dataset of more than 7500 T g values and resulted in a root mean square error (RMSE) of 38.1 °C. The QC-based regression model was trained on 83 T g values and produced an RMSE of 34.5 °C. In conclusion, this work demonstrated that while both model techniques produce accurate predictions and are suitable for different data availability scenarios, the QC-based regression model offered a more interpretable model framework with significantly less training data.
36 MATERIALS SCIENCE↗
CIMantic Graphs
CIMantic Graphs (aka CIM-Graph) is a new python library developed by PNNL to reduce the burden of working with the Common Information Model. CIMantic Graphs takes a novel approach of building in-memory labeled property graphs for creating, parsing, and editing CIM power system models.
Anderson, Alexander↗
Decode the Workload: Training Deep Learning Models for Efficient Compute Cluster Representation
Monitoring the status of a high throughput computing cluster running computationally intensive production jobs is a crucial yet challenging system administration task due to the complexity of such systems. To this end, we train autoencoders using the Linux kernel CPU metrics of the cluster. Additionally, we explore assisting these models with graph neural networks to share information across threads within a compute node. The models are compared in terms of their ability to: 1) Produce a compressed latent representation that captures the salient features of the input, 2) Detect anomalous activity, and 3) Make distinction between different kinds of jobs run at Jefferson Lab. The goal is to have a robust encoder whose compressed embeddings are used for several downstream tasks. We extend this study further by deploying these models in a human-in-the-loop production-based setting for the anomaly detection task and discuss the associated implementation aspects such as continual learning and the criterion to generate alarms. This study represents a first step in the endeavor towards building self-supervised large-scale foundation models for computing centers.
Mohammed, Ahmed↗
Time Series Foundation Models and Deep Learning Architectures for Earthquake Temporal and Spatial Nowcasting
Advancing the capabilities of earthquake nowcasting, the real-time forecasting of seismic activities, remains crucial for reducing casualties. This multifaceted challenge has recently gained attention within the deep learning domain, facilitated by the availability of extensive earthquake datasets. Despite significant advancements, the existing literature on earthquake nowcasting lacks comprehensive evaluations of pre-trained foundation models and modern deep learning architectures; each focuses on a different aspect of data, such as spatial relationships, temporal patterns, and multi-scale dependencies. This paper addresses the mentioned gap by analyzing different architectures and introducing two innovative approaches called Multi Foundation Quake and GNNCoder. We formulate earthquake nowcasting as a time series forecasting problem for the next 14 days within 0.1-degree spatial bins in Southern California. Earthquake time series are generated using the logarithm energy released by quakes, spanning 1986 to 2024. Our comprehensive evaluations demonstrate that our introduced models outperform other custom architectures by effectively capturing temporal-spatial relationships inherent in seismic data. The performance of existing foundation models varies significantly based on the pre-training datasets, emphasizing the need for careful dataset selection. However, we introduce a novel method, Multi Foundation Quake, that achieves the best overall performance by combining a bespoke pattern with Foundation model results handled as auxiliary streams.
97 MATHEMATICS AND COMPUTING↗
Discrete Green’s functions and spectral graph theory for computationally efficient thermal modeling
Here, this work concerns solutions of the heat equation with the spectral graph method, for which the temperature is defined at discrete points in the domain and the spatial relationship among the points is described by a graph. The heat equation on the graph is solved using matrix techniques involving the eigenvectors and eigenvalues of the Laplacian matrix. The spectral graph approach precludes the computationally intensive meshing and numerous time-integration steps of the finite element method. In the present work, the spectral graph method is extended to include heat loss at the boundaries with a generalized boundary condition, and physics-based edge weights are introduced which simplify the calibration process. From this approach a discrete Green’s function is defined which allows for solutions under a variety of heating conditions including: space-varying initial conditions; time-and-space varying internal heating; and, time-and-space-varying heating at boundaries of type 1 (Dirichlet), type 2 (Neumann) and type 3 (Robin). Results are provided for benchmark heat transfer problems in one spatial dimension and in three spatial dimensions, and verification is provided by comparison with exact analytical solutions and finite difference solutions. The spectral graph method converges within 0.4% error of the analytical solution. The practical utility of the approach is demonstrated by thermal simulation of a multilayer additive manufacturing process. The spectral graph results are compared to experimentally-obtained temperature data for two metal parts, with error less than 5% of the experimental measurements, with computation time less than one minute on a desktop computer.
36 MATERIALS SCIENCE↗
Scaling whole-chip QAOA for higher-order ising spin glass models on heavy-hex graphs
Abstract We show that the quantum approximate optimization algorithm (QAOA) for higher-order, random coefficient, heavy-hex compatible spin glass Ising models has strong parameter concentration across problem sizes from 16 up to 127 qubits for p = 1 up to p = 5, which allows for computationally efficient parameter transfer of QAOA angles. Matrix product state (MPS) simulation is used to compute noise-free QAOA performance. Hardware-compatible short-depth QAOA circuits are executed on ensembles of 100 higher-order Ising models on noisy IBM quantum superconducting processors with 16, 27, and 127 qubits using QAOA angles learned from a single 16-qubit instance using the JuliQAOA tool. We show that the best quantum processors find lower energy solutions up to p = 2 or p = 3, and find mean energies that are about a factor of two off from the noise-free distribution. We show that p = 1 QAOA energy landscapes remain very similar as the problem size increases using NISQ hardware gridsearches with up to a 414 qubit processor.
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
Acceleration of Graph Neural Network-Based Prediction Models in Chemistry via Co-Design Optimization on Intelligence Processing Units
Atomic structure prediction and associated property calculations are the bedrock of chemical physics. Since high-fidelity ab initio modeling techniques for computing the structure and properties can be prohibitively expensive, this motivates the development of machine-learning (ML) models that make these predictions more efficiently. Training graph neural networks over large atomistic databases introduces unique computational challenges such as the need to process millions of small graphs with variable size and support communication patterns that are distinct from learning over large graphs such as social networks. We demonstrate a novel hardware-software co-design approach to scale up the training of atomistic graph neural networks (GNN) for structure and property prediction. First, to eliminate redundant computation and memory associated with alternative padding techniques and to improve throughput via minimizing communication, we formulate the effective coalescing of the batches of variable-size atomistic graphs as the bin packing problem and introduce a hardware-agnostic algorithm to pack these batches. In addition, we propose hardware-specific optimizations including a planner and vectorization for the gather-scatter operations targeted for Graphcore’s Intelligence Processing Unit (IPU), as well as model-specific optimizations such as merged communication collectives and optimized softplus. Putting these all together, we demonstrate the effectiveness of the proposed co-design approach by providing an implementation of a well-established atomistic GNN on the Graphcore IPUs. We evaluate the training performance on multiple atomistic graph databases with varying degrees of graph counts, sizes and sparsity. Here, we demonstrate that such a co-design approach can reduce the training time of atomistic GNNs and can improve the performance by up to 1.5× compared to the baseline implementation of the model on the IPUs. Additionally, we compare our IPU implementation with a Nvidia GPU-based implementation and show that our atomistic GNN implementation on the IPUs can run 1.8× faster on average compared to the execution time on the GPUs.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
A graph signal processing‐based multiple model Kalman filter ( GSP‐MMKF ) tool for predictive analytics: An air separation unit process application
Abstract The industrial Air Separations Unit (ASU) is a complicated and tightly operated process. The use of dynamic process analytics is also a key element of safe and economic operation of these processes, with increasing focus on predictive analytics to take preemptive actions. With the availability of real‐time data from hundreds of sensors, the data analysis process should also consider the topology of the data, as seen in sensor networks. In this paper, a novel tool is presented that considers the complex connectivity patterns in the sensor network and uses local adaptive disturbance estimations to predict global network‐scale trends. The paper introduces the emerging field of Graph Signal Processing (GSP) and presents a rigorous derivation of the tool starting from the extraction of the sensor‐network (in a graph theoretical sense) from the data. This network, which is in the form of a matrix, is then used to derive a Kalman‐filter type of state‐space model driven by input disturbances. Multiple disturbance models (e.g., step, ramp, periodic) are included to allow the model to have different kinds of disturbance propagation. Each graph node (representing the sensors used) dynamically adapts to the most recent detected disturbance individually. These estimated disturbances are propagated to the global network using the graph. Modifications to ensure stability are also discussed. The fidelity of the tool is tested on certain downtime events and the paper concludes by discussing the advantages of the method and planned future improvements.
Ghosh, Sambit↗