Amortized Filtering and Smoothing with Conditional Normalizing Flows
This work proposes an amortized framework for filtering and smoothing that reuses trained conditional models across observation sequences and assimilation times and shows that, under the Markov assumption, a summary sufficient for filtering is also sufficient for the backward kernel.
Tiangang Cui, Xiao-Dong Feng, Chenlong Pei et al.
· 2 citations