Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Multilevel CNNs for Parametric PDEs
Cosmas Heiß, Ingo Gühring, Martin Eigel
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 |
|---|---|---|---|
| train ML-Net with halved sample count per finer level instead of the full 1000 samples (ablation) | ML-Net · uniform p=100 | MR H1 error: 2.27e-4 -> 2.86e-4 | No clear effect |
| train ML-Net with halved sample count per finer level instead of the full 1000 samples (ablation) | ML-Net · log-normal p=100 | MR H1 error: 2.64e-4 -> 4.05e-4 | No clear effect |
| use ML-Net on decaying dataset instead of UNetSeq on 232 full-resolution samples | ML-Net · uniform p=100 | MR H1 error: 1.14e-3 -> 2.86e-4 | Improved |
| use ML-Net on decaying dataset instead of UNetSeq on 232 full-resolution samples | ML-Net · log-normal p=100 | MR H1 error: 5.92e-3 -> 4.05e-4 | Improved |
| reduce UNetSeq training set from 1000 to 232 full-resolution samples (ablation) | UNetSeq · log-normal p=100 | MR H1 error: 2.82e-3 -> 5.92e-3 | Hurt |
| use multilevel ML-Net instead of UNetSeq | ML-Net · uniform p=50 | MR H1 error: 2.08e-3 -> 2.21e-4 | Improved |
| use multilevel ML-Net instead of UNetSeq | ML-Net · log-normal p=10 | MR H1 error: 2.83e-3 -> 2.15e-4 | Improved |
| use multilevel ML-Net instead of UNetSeq | ML-Net · uniform p=10 | MR H1 error: 9.75e-5 -> 2.24e-4 | Hurt |
| use multilevel ML-Net instead of UNetSeq | ML-Net · cookie variable p=128 | MR H1 error: 3.08e-3 -> 5.98e-3 | Hurt |
| use multilevel ML-Net instead of UNetSeq | ML-Net · cookie fixed p=64 | MR H1 error: 7.20e-4 -> 1.85e-3 | Hurt |
| increase parameter dimension p from 10 to 200 | ML-Net · uniform | MR H1 error: 2.24e-4 -> 2.31e-4 | No clear effect |
| increase parameter dimension p from 16 to 64 | ML-Net · cookie fixed | MR H1 error: 9.41e-4 -> 1.85e-3 | Hurt |
Closest papers by meaning
- Neural-HSS: Hierarchical Semi-Separable Neural PDE Solver (ICML 2026 · similarity 0.89)
- UGrid: An Efficient-And-Rigorous Neural Multigrid Solver for Linear PDEs (ICML 2024 · similarity 0.89)
- Learning a Neural Solver for Parametric PDEs to Enhance Physics-Informed Methods (ICLR 2025 · similarity 0.87)
- Parameterized Physics-informed Neural Networks for Parameterized PDEs (ICML 2024 · similarity 0.87)
- MultiPDENet: PDE-embedded Learning with Multi-time-stepping for Accelerated Flow Simulation (ICML 2025 · similarity 0.87)
- PIG: Physics-Informed Gaussians as Adaptive Parametric Mesh Representations (ICLR 2025 · 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.
