PapersBuilders12:42Chinese contentNeural CGM:用隐空间条件高斯结构加速非线性数据同化与不确定性量化做气象、流体或地球物理同化的朋友可以看看这篇。思路是把不可观测的物理状态编码进条件高斯隐空间,同化更新有闭式解,比传统非线性滤波省算力,还能直接拿学到的动态做预报,不用已知控制方程。#data assimilation#Neural CGM#uncertainty quantificationaarXiv cs.LG@Chuanqi Chen, Nan ChenSourceRead laterReadWorth followingUsefulFollow data assimilation