Research map / Time series / Causal and generative time series
Probabilistic Weather Forecasting with Deterministic Guidance-based Diffusion Model
Donggeun Yoon, Minseok Seo, Doyi Kim, Yeji Choi, Donghyeon Cho
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 |
|---|---|---|---|
| guide the diffusion model with a deterministic branch (DB+PB) (ablation) | DGDM · Moving MNIST | MAE: 123.20 -> 50.22 | Improved |
| add Brownian bridge to the probabilistic branch (ablation) | DGDM · Moving MNIST | MAE: 110.07 -> 52.04 | Improved |
| use last frame as input to the probabilistic branch (ablation) | DGDM · Moving MNIST | MAE: 52.04 -> 50.35 | Improved |
| add sequential variance schedule (ablation) | DGDM · Moving MNIST | FVD: 9.31 -> 8.28 | Improved |
| add truncated diffusion using deterministic output (200 to 100 steps) (ablation) | DGDM-PB · Moving MNIST | MSE: 23.621 -> 20.963 | Improved |
| reduce denoising steps without truncated diffusion (ablation) | DGDM-PB · Moving MNIST | MAE: 51.223 -> 52.046 | No clear effect |
| train DB and PB jointly (ablation) | DGDM-DB · Moving MNIST | MAE: 58.13 -> 56.45 | Improved |
| add deterministic branch to diffusion model versus deterministic-only TAU | DGDM-PB · Moving MNIST | FVD: 28.169 -> 7.461 | Improved |
| DGDM-PB versus TAU on Moving MNIST MSE | DGDM-PB · Moving MNIST | MSE: 15.68 -> 20.96 | Hurt |
| DGDM versus TAU on WeatherBench wind | DGDM-PB · WeatherBench wind uv10 | MSE: 1.593 -> 1.772 | Hurt |
Closest papers by meaning
- Tyche: One Step Flow for Efficient Probabilistic Weather Forecasting (NeurIPS 2026 · similarity 0.86)
- Continuous Ensemble Weather Forecasting with Diffusion models (ICLR 2025 · similarity 0.86)
- Elucidated Rolling Diffusion Models for Probabilistic Forecasting of Complex Dynamics (NeurIPS 2025 · similarity 0.86)
- Probabilistic Weather Forecasting with Hierarchical Graph Neural Networks (NeurIPS 2024 · similarity 0.85)
- Non-stationary Diffusion For Probabilistic Time Series Forecasting (ICML 2025 · similarity 0.85)
- Fixing the Double Penalty in Data-Driven Weather Forecasting Through a Modified Spherical Harmonic Loss Function (ICML 2025 · similarity 0.84)
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.
