Learning States Representations in POMDP

Gabriella Contardo, Ludovic Denoyer, Thierry Artieres +1

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.