Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Iterative Training of Physics-Informed Neural Networks with Fourier-enhanced Features
Yulun Wu, Miguel Aguiar, Karl Henrik Johansson, Matthieu Barreau
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 two-stage IFeF training instead of end-to-end joint optimization (ablation) | IFeF-PINN · Helmholtz low-frequency | relative L2 error: 0.0088 -> 0.0003 | Improved |
| use two-stage IFeF training instead of end-to-end (ablation) | IFeF-PINN · Viscous Burgers | relative L2 error: 0.0049 -> 0.0024 | Improved |
| use two-stage IFeF training instead of end-to-end (ablation) | IFeF-PINN · Helmholtz high-frequency | relative L2 error: - -> 0.0156 | Improved |
| add RFF basis extension to iterative two-step training (ablation) | IFeF-PINN · Convection high-frequency and Helmholtz high-frequency | relative L2 error | Improved |
| add RFF basis extension to iterative two-step training (ablation) | IFeF-PINN · Convection low-frequency | relative L2 error: 1.4923 x 10^-2 | No clear effect |
| replace Vanilla PINN with IFeF | Vanilla PINN · Convection high-frequency (beta=200) | relative L2 error: 0.9024 -> 0.0027 | Improved |
| replace Vanilla PINN with IFeF | Vanilla PINN · Convection-Diffusion multi-scale | relative L2 error: 0.0501 -> 0.0009 | Improved |
| add primal-dual weight balancing to IFeF | IFeF · Helmholtz high-frequency | relative L2 error: 0.0156 -> 0.0092 | Improved |
| add primal-dual weight balancing to IFeF | IFeF · Viscous Burgers | relative L2 error: 0.0024 -> 0.0033 | Hurt |
| increase Fourier features D from 800 to 2400 (ablation) | IFeF · Helmholtz high-frequency | relative L2 error: 7.11 x 10^-2 -> 1.56 x 10^-2 | Improved |
| decrease sigma from 20 to 10 then to 0.2 (ablation) | IFeF · Helmholtz high-frequency | relative L2 error: 4.6 x 10^-3 -> 3.0 x 10^-3 -> 1.05 x 10^-1 | Mixed |
| vary D in 100 to 3000 (ablation) | IFeF · Helmholtz low-frequency | relative L2 error: 5.5 x 10^-4 -> 2.1 x 10^-4 -> 4.5 x 10^-4 | No clear effect |
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.
- PINNsformer — PINNsFormer: A Transformer-Based Framework For Physics-Informed Neural Networks (ICLR 2024) · better on 2 of 2 shared results
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
- CoPINN: Cognitive Physics-Informed Neural Networks (ICML 2025 · similarity 0.86)
- The Challenges of the Nonlinear Regime for Physics-Informed Neural Networks (NeurIPS 2024 · similarity 0.86)
- Neuro-Spectral Architectures for Causal Physics-Informed Networks (NeurIPS 2025 · similarity 0.86)
- ConFIG: Towards Conflict-free Training of Physics Informed Neural Networks (ICLR 2025 · similarity 0.86)
- Challenges in Training PINNs: A Loss Landscape Perspective (ICML 2024 · similarity 0.86)
- Efficient Error Certification for Physics-Informed Neural Networks (ICML 2024 · 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.
