Research map / Probabilistic and scientific modeling / Bayesian optimization
Meta Learning of Interface Conditions for Multi-Domain Physics-Informed Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 learned interface conditions instead of randomly selecting them for each phase | Multi-domain PINN · Poisson | relative L2 error: 0.3004 ± 0.0032 -> 0.01406 ± 9.1099e-5 | Helped |
| Use learned interface conditions instead of randomly selecting them for each phase | Multi-domain PINN · Advection | relative L2 error: 0.03922 ± 0.00919 -> 0.00070 ± 3.2790e-5 | Helped |
| Use learned interface conditions instead of random single-condition selection | Multi-domain PINN · Reaction | relative L2 error: 0.00612 ± 0.000278 -> 0.00099 ± 5.999e-6 | Helped |
| Use learned interface conditions instead of randomly selecting them for each phase | Multi-domain PINN · Burgers | relative L2 error: 0.03486 ± 0.00066 -> 0.00139 ± 3.948e-5 | Helped |
| Use two sequential interface-condition sets rather than one set for the whole training run (ablation) | METALIC-UCB · Poisson | relative L2 error: 0.0245± 0.0002 -> 0.01406 ± 9.1099e-5 | Helped |
| Use two sequential interface-condition sets rather than one set for the whole training run (ablation) | METALIC-UCB · Advection | relative L2 error: 0.00078 ± 3.6771e-5 -> 0.00070 ± 3.2790e-5 | Helped |
| Use two sequential interface-condition sets rather than one set for the whole training run (ablation) | METALIC-UCB · Reaction | relative L2 error: 0.00102 ± 8.945e-6 -> 0.00099 ± 5.999e-6 | No clear effect |
| Use two sequential interface-condition sets rather than one set for the whole training run (ablation) | METALIC-UCB · Burgers | relative L2 error: 0.00161 ± 2.939e-5 -> 0.00139 ± 3.948e-5 | Hurt |
| Use UCB rather than Thompson sampling for online selection (ablation) | METALIC-Seq · Poisson | relative L2 error: 0.01639 ± 9.5384e-5 -> 0.01406 ± 9.1099e-5 | Helped |
| Use UCB rather than Thompson sampling for online selection (ablation) | METALIC-Seq · Advection | relative L2 error: 0.00078 ± 3.6473e-5 -> 0.00070 ± 3.2790e-5 | Helped |
| Use UCB rather than Thompson sampling for online selection (ablation) | METALIC-Single · Reaction | relative L2 error: 0.00204 ± 1.013e-4 -> 0.00102 ± 8.945e-6 | Helped |
| Use UCB rather than Thompson sampling for online selection (ablation) | METALIC-Single · Burgers | relative L2 error: 0.00109 ± 1.306e-5 -> 0.00161 ± 2.939e-5 | Hurt |
Closest papers by meaning
- Sub-Sequential Physics-Informed Learning with State Space Model (ICML 2025 · similarity 0.89)
- Gradient Alignment in Physics-informed Neural Networks: A Second-Order Optimization Perspective (NeurIPS 2025 · similarity 0.89)
- Learning a Neural Solver for Parametric PDEs to Enhance Physics-Informed Methods (ICLR 2025 · similarity 0.88)
- RoPINN: Region Optimized Physics-Informed Neural Networks (NeurIPS 2024 · similarity 0.88)
- Efficient Error Certification for Physics-Informed Neural Networks (ICML 2024 · similarity 0.88)
- Parameterized Physics-informed Neural Networks for Parameterized PDEs (ICML 2024 · 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.