Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Structure-Preserving Learning Improves Geometry Generalization in Neural PDEs
Benjamin D Shaffer, Shawn Koohy, Brooks Kinch, M. Ani Hsieh, Nathaniel Trask
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 |
|---|---|---|---|
| replace direct regression with Geo-NeW structure-preserving implicit FEM model (ablation) | inducing-point transformer encoder · NS2d-c++ ID, point-cloud input | prediction error: 8.82e-2 -> 6.76e-2 | Improved |
| use mesh-based input encoding instead of point-cloud (ablation) | Geo-NeW · NS2d-c++ ID | prediction error: 6.76e-2 -> 5.17e-2 | Improved |
| use mesh-based input encoding instead of point-cloud (ablation) | baseline inducing-point encoder · NS2d-c++ ID | prediction error: 8.82e-2 -> 7.21e-2 | Improved |
| use Geo-NeW instead of Transolver | geometry extrapolation · Poly-Poisson OOD | normalized L2 error (x1e-2): 7.04 -> 2.14 | Improved |
| use Geo-NeW instead of Transolver | geometry extrapolation · NS2d-c++ OOD | normalized L2 error (x1e-2): 91.4 -> 42.2 | Improved |
| use Geo-NeW instead of strongest baseline | mild extrapolation · NS2d-c++ OOD (mild) | normalized L2 error (x1e-2): 13.1 -> 7.87 | Improved |
| use Geo-NeW instead of Transolver | in-distribution · NS2d-c++ ID | normalized L2 error (x1e-2): 4.04 -> 1.93 | Improved |
| use Geo-NeW instead of Transolver | in-distribution · NS2d-c | normalized L2 error (x1e-2): 0.63 -> 1.10 | Hurt |
| use Geo-NeW instead of LaMO | standard benchmark · NACA (derived quantity) | normalized L2 error (x1e-2): 0.41 -> 0.82 | Hurt |
| use Geo-NeW instead of LaMO | standard benchmark · Pipe | normalized L2 error (x1e-2): 0.38 -> 0.11 | Improved |
| use Geo-NeW instead of LaMO | standard benchmark · Elasticity (derived quantity) | normalized L2 error (x1e-2): 0.50 -> 0.35 | Improved |
| enforce Dirichlet conditions exactly via boundary partitions (ablation) | Geo-NeW · NS2d-c boundary error | normalized L2 boundary error (u): 2.22e-3 -> 0.00 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Transolver — Transolver: A Fast Transformer Solver for PDEs on General Geometries (ICML 2024) · better on 6 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Riemann Tensor Neural Networks: Learning Conservative Systems with Physics-Constrained Networks (ICML 2025 · similarity 0.89)
- Learning a Neural Solver for Parametric PDEs to Enhance Physics-Informed Methods (ICLR 2025 · similarity 0.88)
- A meshfree exterior calculus for generalizable and data-efficient learning of physics from point clouds (NeurIPS 2026 · similarity 0.87)
- AROMA: Preserving Spatial Structure for Latent PDE Modeling with Local Neural Fields (NeurIPS 2024 · similarity 0.87)
- Reference Neural Operators: Learning the Smooth Dependence of Solutions of PDEs on Geometric Deformations (ICML 2024 · similarity 0.87)
- Graph Neural PDE Solvers with Conservation and Similarity-Equivariance (ICML 2024 · similarity 0.87)
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.
