Learning States Representations in POMDP

2013/12/20 by Contardo, Gabriella, Denoyer, Ludovic, Artieres, Thierry +1
#FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.1312.6042

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.

Related

• Toward Learning POMDPs Beyond Full-Rank Actions and State Observability
• Hilbert Space Embeddings of POMDPs
• Embed to Control Partially Observed Systems: Representation Learning with Provable Sample Efficiency
• Agent-state based policies in POMDPs: Beyond belief-state MDPs
• Learning in POMDPs is Sample-Efficient with Hindsight Observability