Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Heavy-tailed Physics-Informed Neural Networks
Jephte Abijuru, Mayank Nagda, Jan Tauberschmidt, Phil Ostheimer, Sebastian Josef Vollmer, Stephan Mandt, Marius Kloft, Sophie Fellenz
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 |
|---|---|---|---|
| replace Gaussian (MSE) residual model with Student-t EM-weighted loss | PINN with MLP 3x32 · Failure-mode benchmarks (Wave, Reaction, Convection, Allen-Cahn) | rRMSE: Allen-Cahn best baseline 0.0545 -> 0.0097; Wave 0.0081 -> 0.0019 | Improved |
| apply t-PINN loss to a standard MLP | MLP PINN · Wave | relative l2 error: 9.0 × 10−3 -> 1.6 × 10−3 | Improved |
| apply t-PINN loss to a standard MLP | MLP PINN · Allen-Cahn | relative l2 error: 2.1 × 10−2 -> 2.7 × 10−3 | Improved |
| apply t-PINN loss to a standard MLP | MLP PINN · Korteweg-de Vries | relative l2 error: 1.6 × 10−1 -> 2.9 × 10−2 | Improved |
| apply t-PINN loss to modified MLP | Modified MLP PINN · Ginzburg-Landau | relative l2 error: 3.1 × 10−2 -> 9.9 × 10−3 | Improved |
| apply t-PINN loss to PirateNet | PirateNet PINN · Allen-Cahn | relative l2 error: 1.4 × 10−5 -> 1.9 × 10−5 | Hurt |
| apply t-PINN loss to PirateNet | PirateNet PINN · Korteweg-de Vries | relative l2 error: 4.3 × 10−4 -> 5.1 × 10−4 | Hurt |
| apply t-PINN loss to PirateNet | PirateNet PINN · Wave | relative l2 error: 7.1 × 10−5 -> 6.9 × 10−5 | No clear effect |
| add Student-t likelihood with EM residual weights (fixed nu) (ablation) | Gaussian PINN · 1D convection (MLP) | relative l2 error: 0.0304 -> 0.0109 | Improved |
| learn degrees of freedom nu instead of fixing it (ablation) | Student-t PINN · 1D convection (MLP) | relative l2 error: 0.0109 -> 0.0028 | Improved |
Closest papers by meaning
- Efficient Error Certification for Physics-Informed Neural Networks (ICML 2024 · similarity 0.90)
- Per-Loss Adapters for Gradient Conflict in Physics-Informed Neural Networks (NeurIPS 2026 · similarity 0.89)
- Gradient Alignment in Physics-informed Neural Networks: A Second-Order Optimization Perspective (NeurIPS 2025 · similarity 0.88)
- PINNacle: A Comprehensive Benchmark of Physics-Informed Neural Networks for Solving PDEs (NeurIPS 2024 · similarity 0.88)
- Challenges in Training PINNs: A Loss Landscape Perspective (ICML 2024 · similarity 0.88)
- Physics-Informed Neural Network Policy Iteration: Algorithms, Convergence, and Verification (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.
