Research map / Graph learning / GNN architectures
Multipole Graph Neural Operator for Parametric Partial Differential Equations
Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Andrew Stuart, Kaushik Bhattacharya, 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 hierarchical multilevel graphs to model long-range interactions (ablation) | MGKN · Darcy flow | testing relative L2 error: 0.1054 -> 0.0568 | Improved |
| Use hierarchical multilevel graphs instead of a single graph | MGKN · Burgers equation | relative L2 error: 0.0367 -> 0.0174 | Improved |
| Use MGKN rather than RBM | Darcy flow · Darcy flow | relative L2 error: 0.0244 vs 0.0416 (s=85); 0.0259 vs 0.0420 (s=421) | Hurt |
| Use MGKN rather than PCA+NN | Burgers equation · Burgers equation | relative L2 error: 0.0397 vs 0.0174 (s=64); 0.0393 vs 0.0364 (s=8192) | Improved |
| Train at one mesh resolution and test at other resolutions | MGKN · Darcy flow | relative L2 error: 0.0579; 0.0590; 0.0589 (train s=61, test s'=61, 121, 241) | No clear effect |
| Add further graph levels while increasing sampled-node hierarchy (ablation) | MGKN · Darcy flow | testing relative L2 error: 0.0744 -> 0.0568 | Improved |
| Increase the number of graph levels from one to four at fixed finest sampling (ablation) | MGKN · Darcy flow | testing relative L2 error: 0.0744 -> 0.0568 | Improved |
| Use orthogonal restriction/prolongation transitions rather than overlapping general convolutions | MGKN · Darcy flow and Burgers equation | relative L2 error | Mixed |
Closest papers by meaning
- Graph Neural PDE Solvers with Conservation and Similarity-Equivariance (ICML 2024 · similarity 0.90)
- Physics-Embedded Neural Networks: Graph Neural PDE Solvers with Mixed Boundary Conditions (NeurIPS 2022 · similarity 0.89)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.87)
- RIGNO: A Graph-based Framework For Robust And Accurate Operator Learning For PDEs On Arbitrary Domains (NeurIPS 2025 · similarity 0.86)
- HAMLET: Graph Transformer Neural Operator for Partial Differential Equations (ICML 2024 · similarity 0.86)
- PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systems (ICLR 2025 · similarity 0.86)
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.
