Goal-conditioned Offline Reinforcement Learning through State Space Partitioning

arXiv:2303.09367v2 Announce Type: replace
Abstract: Offline reinforcement learning (RL) aims to infer sequential decision policies using only offline datasets. This is a particularly difficult setup, especially when learning to achieve multiple different goals or outcomes under a given scenario with only sparse rewards. For offline learning of goal-conditioned policies via supervised learning, previous work has shown that an advantage weighted log-likelihood loss guarantees monotonic policy improvement. In this work we argue that, despite its benefits, this approach is still insufficient …

abstract arxiv cs.lg datasets decision multiple offline partitioning policies reinforcement reinforcement learning replace setup space state supervised learning through type via