Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Deep Coupling Learning for Solving PDEs
Lingshi MENG, Haosen Shi, Sinno Jialin Pan
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 permutation layers (ablation) | CoupledNet · High-frequency PDE (6) | relative L2 error: 0.063 -> 0.0117 | Improved |
| deepen CoupledNet from 4 to 8 layers (ablation) | CoupledNet · High-frequency PDE (6) | relative L2 error: 0.0200 -> 0.0117 | Improved |
| deepen CoupledNet from 8 to 24 layers (ablation) | CoupledNet · High-frequency PDE (6) | relative L2 error: 0.0117 -> 0.0116 | No clear effect |
| deepen CoupledNet from 8 to 16 layers | CoupledNet · High-dynamic-range PDE (7) | relative L2 error: 0.5775 -> 0.0718 | Improved |
| deepen PirateNet from 9 to 18 layers | PirateNet · High-dynamic-range PDE (7) | relative L2 error: 2.673 -> 4.067 | Hurt |
| use CoupledNet instead of PirateNet | CoupledNet · High-frequency PDE (6) | relative L2 error: 0.0334 -> 0.0117 | Improved |
| use CoupledNet instead of PirateNet at dimension 40 | CoupledNet · HJB dim 40 | relative L2 error: 0.400 -> 0.137 | Improved |
| use CoupledNet instead of MLP at dimension 40 | CoupledNet · HJB dim 40 | relative L2 error: 0.240 -> 0.137 | Improved |
| use CoupledNet instead of PirateNet on Allen-Cahn | CoupledNet · Jaxpi benchmark: Allen-Cahn | relative L2 error: 2.24e-5 -> 9.09e-5 | Hurt |
| use CoupledNet instead of PirateNet on Advection | CoupledNet · Jaxpi benchmark: Advection | relative L2 error: 4.88e-4 -> 3.64e-4 | Improved |
| use CoupledNet instead of PirateNet on Grey-Scott | CoupledNet · Jaxpi benchmark: Grey-Scott | relative L2 error: 3.61e-3 -> 1.93e-2 | Hurt |
| add NTK reweighting to PirateNet | PirateNet · High-frequency PDE (6) | relative L2 error: 0.0785 -> 0.0334 | Improved |
Closest papers by meaning
- PINNsFormer: A Transformer-Based Framework For Physics-Informed Neural Networks (ICLR 2024 · similarity 0.89)
- Efficient Error Certification for Physics-Informed Neural Networks (ICML 2024 · similarity 0.89)
- Challenges in Training PINNs: A Loss Landscape Perspective (ICML 2024 · similarity 0.89)
- PINNacle: A Comprehensive Benchmark of Physics-Informed Neural Networks for Solving PDEs (NeurIPS 2024 · similarity 0.88)
- Dual Cone Gradient Descent for Training Physics-Informed Neural Networks (NeurIPS 2024 · similarity 0.88)
- Coupled Integral PINN for Discontinuity (NeurIPS 2026 · 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.
