Research map / Label-efficient and robust learning / Test-time adaptation
Buckingham $\pi$-Invariant Test‑Time Projection for Robust PDE Surrogate Modeling
Seokki Lee, Min-Chul Park, Giyong Hong, Changwook Jeong
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 pi-invariant pairwise projection at test time | CNN · 2D thermal conduction (OOD) | MAE: 8.43 -> 2.63 | Improved |
| add pi-invariant pairwise projection at test time | U-Net · 2D thermal conduction (OOD) | MAE: 13.60 -> 1.75 | Improved |
| add pi-invariant pairwise projection at test time | FNO · 2D thermal conduction (OOD) | MAE: 9.88 -> 1.38 | Improved |
| add pi-invariant pairwise projection at test time | FNO · 2D linear elasticity (OOD) | MAE: 3.20 -> 0.28 | Improved |
| replace pairwise projection with pi-uniform + 10 K-means centroids (ablation) | CNN · 2D thermal conduction (OOD) | projection time: 100.31 -> 1.80 | Improved |
| replace pairwise projection with pi-uniform + 10 K-means centroids (ablation) | U-Net · 2D thermal conduction (OOD) | MAE: 1.75 -> 1.18 | Improved |
| use K-means centroids instead of random training samples as projection candidates (ablation) | FNO · 2D thermal conduction (OOD) | MAE: 1.29 -> 1.25 | No clear effect |
| use K-means centroids instead of random training samples as projection candidates (ablation) | CNN · 2D thermal conduction (OOD) | MAE: 1.85 -> 1.79 | No clear effect |
| replace pairwise projection with pi-uniform + 10 centroids (ablation) | CNN · 2D linear elasticity (OOD) | MAE: 0.53 -> 0.60 | Hurt |
| apply projection when the governing pi-group is incomplete (f nonzero) | U-Net · Navier-Stokes (OOD, f≠0, incomplete π) | MAE: 0.105 -> 0.061 | Improved |
| apply projection to extreme OOD Reynolds range | CNN · Navier-Stokes Test-OOD (extreme, Re 42.28-63.43) | MAE: 0.38 -> 0.34 | Improved |
| apply projection to mid-range OOD Reynolds | CNN · Navier-Stokes Test-OOD (mid, Re 0.32-27.06) | MAE: 0.20 -> 0.08 | Improved |
Closest papers by meaning
- Compute-optimal data scaling for neural surrogates via multi-fidelity training (NeurIPS 2026 · similarity 0.85)
- Geometric and Physical Constraints Synergistically Enhance Neural PDE Surrogates (ICML 2025 · similarity 0.85)
- Efficient Learning of PDEs via Taylor Expansion and Sparse Decomposition into Value and Fourier Domains (AAAI 2024 · similarity 0.84)
- Active Learning with Selective Time-Step Acquisition for PDEs (ICML 2025 · similarity 0.84)
- Uncertainty Quantification for Forward and Inverse Problems of PDEs via Latent Global Evolution (AAAI 2024 · similarity 0.83)
- Physics-informed Reduced Order Modeling of Time-dependent PDEs via Differentiable Solvers (NeurIPS 2025 · similarity 0.83)
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.
