Work

Curriculum learning for mesh-based simulations

Year: 2026

Type: article

Abstract: Graph neural networks (GNNs) have emerged as powerful surrogates for mesh-based computational fluid dynamics, but training them on high-resolution unstructured meshes with hundreds of thousands of nodes remains prohibitively expensive. We study a coarse-to-fine curriculum that accelerates convergence by first training on very coarse meshes and then progressively introducing medium and high resolutions (up to 3×105 nodes). Unlike multiscale GNN architectures, the model itself is unchanged; only the fidelity of the training data varies over time. We achieve comparable generalization accuracy while reducing total wall-clock time by up to 50%. Furthermore, on datasets where our model lacks the capacity to learn the underlying physics, using curriculum learning enables it to break through plateaus.

Source: Physics of Fluids

DOI: 10.1063/5.0316165

Authors: Paul Garnier, Vincent Lannelongue, Elie Hachem

Institutions: Centre National de la Recherche Scientifique, Université Paris Sciences et Lettres, Centre de Mise en Forme des Matériaux

Language: English

Topic: Model Reduction and Neural Networks

Subfield: Statistical and Nonlinear Physics

Field: Physics and Astronomy

Domain: Physical Sciences

SDG: Quality education

Open Access status: green

Funder: HORIZON EUROPE European Research Council