Keywords: imitation learning, reinforcement learning, expert data, hidden confounding, causal inference, covariate shift
Abstract: We consider the problem of using expert data with unobserved confounders for imitation and reinforcement learning. We begin by defining the problem of learning from confounded expert data in a contextual MDP setup. We analyze the limitations of learning from such data with and without external reward and propose an adjustment of standard imitation learning algorithms to fit this setup. In addition, we discuss the problem of distribution shift between the expert data and the online environment when partial observability is present in the data. We prove possibility and impossibility results for imitation learning under arbitrary distribution shift of the missing covariates. When additional external reward is provided, we propose a sampling procedure that addresses the unknown shift and prove convergence to an optimal solution. Finally, we validate our claims empirically on challenging assistive healthcare and recommender system simulation tasks.
One-sentence Summary: We use expert data with unobserved confounders for both imitation and reinforcement learning. Such hidden confounding is prone to a shifted distribution, which may severely hurt performance unless accounted for.
Supplementary Material: zip