Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Geometry-Informed Neural Operator for Large-Scale 3D PDEs
Zongyi Li, Nikola Borislavov Kovachki, Chris Choy, Boyi Li, Jean Kossaifi, Shourya Prakash Otta, Mohammad Amin Nabian, Maximilian Stadler, Christian Hundt, Kamyar Azizzadenesheli, Anima Anandkumar
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 |
|---|---|---|---|
| use GINO (decoder) instead of best prior baseline UNet (interp) | GINO · Ahmed-body | test error: 11.16% -> 8.31% | Improved |
| use GINO (decoder) instead of FNO (interp) | GINO · Shape-Net Car | test error: 9.42% -> 7.12% | Improved |
| add GNO encoder to GINO (encoder-decoder vs decoder-only) (ablation) | GINO · Shape-Net Car | test error: 7.12% -> 9.47% | No clear effect |
| add GNO encoder to GINO (encoder-decoder vs decoder-only) (ablation) | GINO · Ahmed-body | test error: 8.31% -> 9.01% | No clear effect |
| increase latent resolution from 32 to 80 at fixed radius 0.055 (ablation) | GINO · Ahmed-body | test error: 13.59% -> 7.87% | Improved |
| scale radius with latent resolution to fix number of edges (32 to 80) (ablation) | GINO · Ahmed-body | test error: 10.10% -> 7.89% | Improved |
| increase GNO radius from 0.025 to 0.035 (ablation) | GINO (decoder) · Ahmed-body | test error: 12.74% -> 9.39% | Improved |
| add Riemann-sum weights in the kernel of the encoder (weighted) (ablation) | GINO (encoder-decoder) · Ahmed-body radius 0.035 | test error: 10.32% -> 9.90% | Improved |
| add Riemann-sum weights in the kernel of the encoder (weighted) (ablation) | GINO (encoder-decoder) · Ahmed-body radius 0.025 | test error: 13.07% -> 12.76% | Improved |
| train on sub-sampled mesh (1/2 to 1/8) and test on full mesh (zero-shot super-resolution) (ablation) | GINO (encoder-decoder) · Ahmed-body | test error: 8.8% -> 9.7% | No clear effect |
| train UNet (interp) on sub-sampled mesh and test on full mesh | UNet (interp) · Ahmed-body | error: 16.5% -> 15.6% | Mixed |
| use a larger UNet (more channels/depth) | 3D UNet · Ahmed-body | test error: 11.1% | Improved |
Closest papers by meaning
- AeroGTO: An Efficient Graph-Transformer Operator for Learning Large-Scale Aerodynamics of 3D Vehicle Geometries (AAAI 2025 · similarity 0.89)
- Geometry Aware Operator Transformer as an efficient and accurate neural surrogate for PDEs on arbitrary domains (NeurIPS 2025 · similarity 0.88)
- RIGNO: A Graph-based Framework For Robust And Accurate Operator Learning For PDEs On Arbitrary Domains (NeurIPS 2025 · similarity 0.88)
- Reference Neural Operators: Learning the Smooth Dependence of Solutions of PDEs on Geometric Deformations (ICML 2024 · similarity 0.87)
- GAIA: Geometry-Adaptive Operator Learning for Forward and Inverse Problems (NeurIPS 2026 · similarity 0.87)
- BENO: Boundary-embedded Neural Operators for Elliptic PDEs (ICLR 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.
