Learning States Representations in POMDP

Abstract

We propose to deal with sequential processes where only partial observations are available by learning a latent representation space on which policies may be accurately learned.

arXiv comment4 pages