Interpretable Fine-Tuning for Graph Neural Network Surrogate Models

arXiv:2311.07548v2 Announce Type: replace
Abstract: Data-driven surrogate modeling has surged in capability in recent years with the emergence of graph neural networks (GNNs), which can operate directly on mesh-based representations of data. The goal of this work is to introduce an interpretable fine-tuning strategy for GNNs, with application to unstructured mesh-based fluid dynamics modeling. The end result is an enhanced fine-tuned model that isolates regions in physical space, corresponding to sub-graphs, that are intrinsically linked to the forecasting task while …

abstract application arxiv capability cs.lg data data-driven emergence fine-tuning gnns graph graph neural network graph neural networks mesh modeling network networks neural network neural networks physics.comp-ph physics.flu-dyn strategy type unstructured work