Research map / Graph learning / Knowledge and temporal graphs
Mesh Based Simulations with Spatial and Temporal awareness
Paul Garnier, Vincent Lannelongue, Elie Hachem
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Multi Node Prediction auxiliary task (stencil-level loss) (ablation) | Transformer · Aneurysm | 1-step and all-rollout RMSE | Improved |
| increase MNP centers from 256 to 1024 (ablation) | Transformer + MNP · Aneurysm | RMSE | Improved |
| replace MNP loss with an L2 loss on the field gradient (ablation) | Transformer · Aneurysm | RMSE | Hurt |
| bias MNP center sampling toward or away from contact region instead of uniform (ablation) | MNP · DeformingPlate | RMSE | No clear effect |
| use cross-attention only, without gating and mixing, as temporal correction (ablation) | Transformer · Aneurysm | RMSE | Hurt |
| add gated temporal correction with mixing MLP (ablation) | Transformer · Aneurysm | 1-step and all-rollout RMSE | Improved |
| apply temporal correction after every spatial block vs only the last (ablation) | Transformer · Aneurysm | RMSE | Improved |
| add intermediate supervised loss after each temporal correction (ablation) | Transformer + temporal correction · Aneurysm | RMSE | Hurt |
| add 3D RoPE in place of absolute coordinates in node features (ablation) | Transformer · Aneurysm | RMSE | No clear effect |
| use 3D RoPE together with absolute coordinates (ablation) | Transformer · Aneurysm | RMSE | Improved |
| use learned positional embedding, learnable relative bias or weighted adjacency instead of RoPE (ablation) | Transformer · Aneurysm | RMSE | Hurt |
| match parameter count of baseline by widening instead of adding the proposed modules (ablation) | MeshGraphNet, Transolver, Transformer · Cylinder, Aneurysm, DeformingPlate | RMSE | No clear effect |
Closest papers by meaning
- Learning Distributions of Complex Fluid Simulations with Diffusion Graph Networks (ICLR 2025 · similarity 0.91)
- MeshMask: Physics-Based Simulations with Masked Graph Neural Networks (ICLR 2025 · similarity 0.90)
- UGM2N: An Unsupervised and Generalizable Mesh Movement Network via M-Uniform Loss (NeurIPS 2025 · similarity 0.89)
- PIORF: Physics-Informed Ollivier-Ricci Flow for Long–Range Interactions in Mesh Graph Neural Networks (ICLR 2025 · similarity 0.89)
- Space and time continuous physics simulation from partial observations (ICLR 2024 · similarity 0.89)
- Finite Volume Features, Global Geometry Representations, and Residual Training for Deep Learning-based CFD Simulation (ICML 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
