arXiv：cs.LG· Jost Arndt, Utku Isil, Noelia Otero, Rodrigo Almeida, Wojciech Samek, Jackie Ma·· 3 小时前AI 评分32
SoftSEEPS：可微分 SEEPS 评分让机器学习直接预报降水
SoftSEEPS improves ML-based precipitation forecasting
AI 导读
研究者提出 SoftSEEPS，即经典 SEEPS 评分的可微分近似版本，使机器学习模型能够直接以降水预报为目标进行训练。他们在 0.1 度分辨率的 IMERG 数据集上，于预训练低分辨率预报模型的隐空间上训练降水解码器进行验证。将 SoftSEEPS 与 RMSE 组合为联合目标可行，两项指标仅有轻微取舍。
正文
Abstract:In this paper we have developed a differentiable approximation of the well-known SEEPS score, which we name SoftSEEPS. This allows the training of a Machine Learning model to forecast precipitation directly. We test SoftSEEPS on the IMERG dataset (0.1 degree resolution) by training a decoder for precipitation on the latent space of a pre-trained low-resolution forecasting model. Combining SoftSEEPS and RMSE in a joint objective is possible with marginal trade-offs in either metric.
| Subjects: | Machine Learning (cs.LG) |
| ACM classes: | I.2.1 |
| Cite as: | arXiv:2610.09752 [cs.LG] |
| (or arXiv:2610.09752v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.09752 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jost Arndt [view email]
[v1]
Wed, 7 Oct 2026 09:40:30 UTC (2,479 KB)
来源：arXiv：cs.LG · arxiv.org