Regret Bounds for Risk-Sensitive Reinforcement Learning

10/11/2022
by   O. Bastani, et al.
0

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.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset