Provably efficient multi-task reinforcement learning with model transfer

Chicheng Zhang, Zhi Wang

Keywords: Multi-task learning, Provably efficient reinforcement learning, Model transfer

TL;DR: We formulate a new multi-task episodic reinforcement learning problem, and provide a provably efficient algorithm using model transfer in this setting.

Abstract: We study multi-task reinforcement learning (RL) in tabular episodic Markov decision processes (MDPs). We formulate a heterogeneous multi-player RL problem, in which a group of players concurrently face similar but not necessarily identical MDPs, with a goal of improving their collective performance through inter-player information sharing. We design and analyze a model-based algorithm, and provide gap-dependent and gap-independent regret upper and lower bounds that characterize the intrinsic complexity of the problem.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.