Research map / Probabilistic and scientific modeling / Dynamical systems
DYffusion: A Dynamics-informed Diffusion Model for Spatiotemporal Forecasting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 DYffusion instead of Dropout ensemble baseline | Navier-Stokes · Navier-Stokes full trajectory | CRPS: 0.078 -> 0.067 | Helped |
| use DYffusion instead of MCVD | SST · SST 7-day forecast | CRPS: 0.216 -> 0.224 | Hurt |
| use DYffusion instead of MCVD | SST · SST 7-day forecast | sampling time (s): 79.167 -> 4.6722 | Helped |
| reduce training data to 2 years (DYffusion vs MCVD) | DYffusion · SST CRPS | CRPS: 0.262 -> 0.239 | Helped |
| use DYffusion instead of Dropout | Spring Mesh OOD · Spring Mesh out-of-distribution | CRPS: 0.0448 -> 0.0292 | Helped |
| disable stochasticity (MC dropout) in the interpolator (ablation) | DYffusion · all datasets | CRPS | Hurt |
| use cold sampling instead of naive sampling (ablation) | DYffusion · SST | CRPS | Helped |
| add auxiliary non-integer diffusion steps (k) (ablation) | DYffusion · SST | CRPS | Helped |
| skip intermediate diffusion steps at inference (ablation) | DYffusion · SST | accuracy vs speed | Mixed |
| multi-step training instead of single-step (h=1) barebone (ablation) | backbone · all datasets | forecast skill | Helped |
Closest papers by meaning
- Elucidated Rolling Diffusion Models for Probabilistic Forecasting of Complex Dynamics (NeurIPS 2025 · similarity 0.90)
- Dynamical Diffusion: Learning Temporal Dynamics with Diffusion Models (ICLR 2025 · similarity 0.90)
- Rolling Diffusion Models (ICML 2024 · similarity 0.89)
- Non-stationary Diffusion For Probabilistic Time Series Forecasting (ICML 2025 · similarity 0.88)
- DiffPTS: Rethinking Diffusion ELBO for Probabilistic Time Series Forecasting (NeurIPS 2026 · similarity 0.88)
- Probablistic Emulation of a Global Climate Model with Spherical DYffusion (NeurIPS 2024 · 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.