Metraipolis › archive › dc_f724b69010dd
Diffusion Distillation for Efficient Weather Ensembles — arXiv:2608.27728
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.27728
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T11:06:26.122Z
- SHA-256 of the archived text
a6e4c07cd642925d633c813fe0ad0e88e59c58228576badaee568134def120bc- Stable URL of this page
https://metraipolis.com/documento/dc_f724b69010dd
How to cite
Diffusion Distillation for Efficient Weather Ensembles — arXiv:2608.27728 — arXiv (en), https://arxiv.org/abs/2608.27728 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f724b69010dd (Metraipolis archive dc_f724b69010dd, sha256 a6e4c07cd642925d633c813fe0ad0e88e59c58228576badaee568134def120bc)
BibTeX
@misc{dcf724b69010dd,
title = {Diffusion Distillation for Efficient Weather Ensembles — arXiv:2608.27728},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.27728},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f724b69010dd; sha256 a6e4c07cd642925d633c813fe0ad0e88e59c58228576badaee568134def120bc},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27728. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Diffusion models generate skillful weather ensembles but require costly iterative sampling. We introduce a supervised energy-distance distillation method that compresses a multi-step diffusion teacher into a single-step student by aligning student forecasts with teacher samples and ground-truth observations. Experiments on global forecasting and typhoon-track prediction show that our student outperforms existing distillation methods and preserves skill for extreme events. It matches or surpasses the teacher across key metrics using only one neural function evaluation per autoregressive step. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-31 in cs.LG). Categorie arXiv: Machine Learning (cs.LG) ; Applications (stat.AP). Pagina del lavoro: https://arxiv.org/abs/2608.27728 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-17T10:50:25.775Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-31?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).