English

Provable Reinforcement Learning with a Short-Term Memory

Machine Learning 2022-02-09 v1 Artificial Intelligence

Abstract

Real-world sequential decision making problems commonly involve partial observability, which requires the agent to maintain a memory of history in order to infer the latent states, plan and make good decisions. Coping with partial observability in general is extremely challenging, as a number of worst-case statistical and computational barriers are known in learning Partially Observable Markov Decision Processes (POMDPs). Motivated by the problem structure in several physical applications, as well as a commonly used technique known as "frame stacking", this paper proposes to study a new subclass of POMDPs, whose latent states can be decoded by the most recent history of a short length mm. We establish a set of upper and lower bounds on the sample complexity for learning near-optimal policies for this class of problems in both tabular and rich-observation settings (where the number of observations is enormous). In particular, in the rich-observation setting, we develop new algorithms using a novel "moment matching" approach with a sample complexity that scales exponentially with the short length mm rather than the problem horizon, and is independent of the number of observations. Our results show that a short-term memory suffices for reinforcement learning in these environments.

Keywords

Cite

@article{arxiv.2202.03983,
  title  = {Provable Reinforcement Learning with a Short-Term Memory},
  author = {Yonathan Efroni and Chi Jin and Akshay Krishnamurthy and Sobhan Miryoosefi},
  journal= {arXiv preprint arXiv:2202.03983},
  year   = {2022}
}
R2 v1 2026-06-24T09:26:43.256Z