Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Physics-Informed Diffusion Models in Spectral Space
Davide Gallon, Philippe von Wurstemberger, Patrick Cheridito, Arnulf Jentzen
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 frequency-aware Adam instead of plain GD in guidance (ablation) | PISD · Inverse Poisson (500 obs) | relative error (%): 49.25 -> 13.81 | Improved |
| use frequency-aware Adam instead of plain GD in guidance (ablation) | PISD · Forward Poisson (500 obs) | relative error (%): 3.60 -> 3.08 | Improved |
| use Adam instead of GD in guidance (ablation) | FunDPS · Inverse Poisson (500 obs) | relative error (%): 21.09 -> 17.44 | Improved |
| use Momentum instead of GD in guidance (ablation) | FunDPS · Inverse Poisson (500 obs) | relative error (%): 21.09 -> 18.93 | Improved |
| add frequency-aware learning rate to Adam (ablation) | PISD · Inverse Poisson (500 obs) | relative error (%): 17.38 -> 13.81 | Improved |
| add frequency-aware learning rate to Adam (ablation) | PISD · Forward Poisson (500 obs) | relative error (%): 3.82 -> 3.08 | Improved |
| add frequency-aware learning rate to Momentum (ablation) | PISD · Inverse Poisson (500 obs) | relative error (%): 42.74 -> 50.10 | Hurt |
| use hyperbolic instead of square truncation at 44x44 (ablation) | PISD · Inverse Poisson (500 obs) | relative error (%): 28.94 -> 13.81 | Improved |
| increase hyperbolic modes from 44x44 to 64x64 (ablation) | PISD · Inverse Poisson (500 obs) | relative error (%): 13.81 -> 12.03 | No clear effect |
| reduce hyperbolic modes from 44x44 to 32x32 (ablation) | PISD · Inverse Poisson (500 obs) | relative error (%): 13.81 -> 20.05 | Hurt |
| vary guidance weight lambda_a over 10^-4 to 10^0 (ablation) | PISD · Forward Poisson (500 obs) | relative error (%): 2.96 to 3.96 | No clear effect |
| use PISD instead of DiffusionPDE/FunDPS | DiffusionPDE · Inverse Poisson (500 obs) | relative error (%): 22.17 -> 13.81 | Improved |
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.
- DiffusionPDE — DiffusionPDE: Generative PDE-Solving under Partial Observation (NeurIPS 2024) · better on 12 of 12 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
- Ultra Fast PDE Solving via Physics Guided Few-step Diffusion (NeurIPS 2026 · similarity 0.90)
- DiffusionPDE: Generative PDE-Solving under Partial Observation (NeurIPS 2024 · similarity 0.89)
- Text2PDE: Latent Diffusion Models for Accessible Physics Simulation (ICLR 2025 · similarity 0.89)
- Physics-Informed Diffusion Models (ICLR 2025 · similarity 0.89)
- Physics-informed Reduced Order Modeling of Time-dependent PDEs via Differentiable Solvers (NeurIPS 2025 · similarity 0.88)
- Holistic Physics Solver: Learning PDEs in a Unified Spectral-Physical Space (ICML 2025 · similarity 0.87)
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.
