{"ID":2870477,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2509.13138","arxiv_id":"2509.13138","title":"Curriculum Learning for Mesh-based simulations","abstract":"Graph neural networks (GNNs) have emerged as powerful surrogates for mesh-based computational fluid dynamics (CFD), but training them on high-resolution unstructured meshes with hundreds of thousands of nodes remains prohibitively expensive. We study a \\emph{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\\times10^5\\) 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.","short_abstract":"Graph neural networks (GNNs) have emerged as powerful surrogates for mesh-based computational fluid dynamics (CFD), but training them on high-resolution unstructured meshes with hundreds of thousands of nodes remains prohibitively expensive. We study a \\emph{coarse-to-fine curriculum} that accelerates convergence by fi...","url_abs":"https://arxiv.org/abs/2509.13138","url_pdf":"https://arxiv.org/pdf/2509.13138v1","authors":"[\"Paul Garnier\",\"Vincent Lannelongue\",\"Elie Hachem\"]","published":"2025-09-16T14:54:11Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"physics.comp-ph\"]","methods":"[\"Graph Neural Network\"]","has_code":false}