Scalable uncertainty quantification for extreme weather forecasting via NTK

Arxiv CS.LG (machine learning)

Paper introduces a scalable uncertainty quantification method using empirical neural tangent kernels for deep learning weather models. The approach addresses the critical lack of uncertainty estimates in deterministic AI weather forecasts for high-stakes extreme weather events.

neural-tangent-kernel

Arxiv CS.LG (machine learning)·Jose Marie Antonio Mi\~nozaRex Gregor LayloSebastian C. Iba\~nez

8 days ago