Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Physics-informed Neural Operator Learning for Nonlinear Grad-Shafranov Equation
Siqi Ding, Zitong Zhang, Shi Guoyang, Xingyu Li, Xiang Gu, Yanan Xu, Huasheng Xie, Hanyue Zhao, YUEJIANG SHI, Tianyuan Liu
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 KAN regressor instead of coordinate MLP (ablation) | coordinate network · GSE pilot 4D IID test | Mean L2(psi) error (%): 4.93 -> 0.34 | Improved |
| use Transformer trunk instead of MLP, FNO or CNN trunk with KAN regressor (ablation) | TKNO · GSE pilot 4D IID test | Mean L2(psi) error (%): 0.54 / 0.36 / 0.30 -> 0.25 | Improved |
| CNN trunk gives low pointwise error but large PDE residual (ablation) | C-K · GSE pilot 4D IID test | MAE(PDE): 1.23 -> 5.23 (vs TKNO) | Hurt |
| train with physics-only losses instead of data-only (ablation) | TKNO · GSE pilot 4D | L2(psi) error (%): 0.25 -> 0.62 | Hurt |
| increase sparse anchors from 10 to 100 points (ablation) | Physics-anchored TKNO · GSE pilot 4D | L2(psi) error (%): 0.75 -> 0.59 | Improved |
| increase sparse anchors from 100 to 1000 points (ablation) | Physics-anchored TKNO · GSE pilot 4D | MSE(PDE): L2 0.59 -> 0.52; MSE(PDE) 0.011 -> 0.023 | Mixed |
| add data anchors to PDE constraints (physics-anchored vs physics-only) (ablation) | TKNO · GSE 10D joint OOD | Mean L2(psi) error (%): 42.77 -> 11.44 | Improved |
| add PDE constraints to data supervision (physics-anchored vs data-only) (ablation) | TKNO · GSE 10D Joint Any-OOD | p99 L2(psi) error (%): 53.66 -> 44.12 | Improved |
| add PDE constraints to data supervision (ablation) | TKNO · GSE 10D isolated Shape OOD | p99 L2(psi) error (%): 52.25 -> 33.16 | Improved |
| add PDE constraints to data supervision (ablation) | TKNO · GSE 10D IID holdout | Mean L2(psi) error (%): 0.24 -> 0.59 | Hurt |
| add PDE constraints to data supervision (ablation) | TKNO · GSE 10D isolated Profile OOD | Mean L2(psi) error (%): 5.80 -> 7.69 | Hurt |
| add data supervision to physics-only training (ablation) | TKNO · GSE 10D joint OOD | branch-scale failure rate: approximately 12% -> below 0.2% | 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 1 of 1 shared result
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
- Learning a Neural Solver for Parametric PDEs to Enhance Physics-Informed Methods (ICLR 2025 · similarity 0.85)
- RIGNO: A Graph-based Framework For Robust And Accurate Operator Learning For PDEs On Arbitrary Domains (NeurIPS 2025 · similarity 0.84)
- Physics-Informed Deep Inverse Operator Networks for Solving PDE Inverse Problems (ICLR 2025 · similarity 0.84)
- Universal Physics Transformers: A Framework For Efficiently Scaling Neural Operators (NeurIPS 2024 · similarity 0.84)
- Model-Agnostic Knowledge Guided Correction for Improved Neural Surrogate Rollout (ICLR 2025 · similarity 0.84)
- Harnessing the Power of Neural Operators with Automatically Encoded Conservation Laws (ICML 2024 · similarity 0.84)
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.
