Efficient Policy Learning for Non-Stationary MDPs under Adversarial Manipulation
share
A Markov Decision Process (MDP) is a popular model for reinforcement learning. However, its commonly used assumption of stationary dynamics and rewards is too stringent and fails to hold in adversarial, nonstationary, or multi-agent problems. We study an episodic setting where the parameters of an MDP can differ across episodes. We learn a reliable policy of this potentially adversarial MDP by developing an Adversarial Reinforcement Learning (ARL) algorithm that reduces our MDP to a sequence of adversarial bandit problems. ARL achieves O(√(SATH^3)) regret, which is optimal with respect to S, A, and T, and its dependence on H is the best (even for the usual stationary MDP) among existing model-free methods.
READ FULL TEXT