Paper ID: 2210.05650

Regret Bounds for Risk-Sensitive Reinforcement Learning

O. Bastani, Y. J. Ma, E. Shen, W. Xu

In safety-critical applications of reinforcement learning such as healthcare and robotics, it is often desirable to optimize risk-sensitive objectives that account for tail outcomes rather than expected reward. We prove the first regret bounds for reinforcement learning under a general class of risk-sensitive objectives including the popular CVaR objective. Our theory is based on a novel characterization of the CVaR objective as well as a novel optimistic MDP construction.

Submitted: Oct 11, 2022