Confounding Robust Deep Reinforcement Learning: A Causal Approach

TL;DR

This paper introduces a new deep reinforcement learning algorithm designed to be robust against unobserved confounding biases, ensuring safer policy learning in complex environments with biased data.

Contribution

It proposes a novel confounding-robust deep Q-learning algorithm that finds safe policies under worst-case confounding scenarios, extending standard DQN methods.

Findings

Outperforms standard DQN in confounded Atari games

Consistently finds safer policies in biased data settings

Effective in high-dimensional, complex environments

Abstract

A key task in Artificial Intelligence is learning effective policies for controlling agents in unknown environments to optimize performance measures. Off-policy learning methods, like Q-learning, allow learners to make optimal decisions based on past experiences. This paper studies off-policy learning from biased data in complex and high-dimensional domains where \emph{unobserved confounding} cannot be ruled out a priori. Building on the well-celebrated Deep Q-Network (DQN), we propose a novel deep reinforcement learning algorithm robust to confounding biases in observed data. Specifically, our algorithm attempts to find a safe policy for the worst-case environment compatible with the observations. We apply our method to twelve confounded Atari games, and find that it consistently dominates the standard DQN in all games where the observed input to the behavioral and target policies mismatch and unobserved confounders exist.