Minimax Regret Bounds for Reinforcement Learning

03/16/2017
by   Mohammad Gheshlaghi Azar, et al.
0

We consider the problem of provably optimal exploration in reinforcement learning for finite horizon MDPs. We show that an optimistic modification to value iteration achieves a regret bound of Õ( √(HSAT) + H^2S^2A+H√(T)) where H is the time horizon, S the number of states, A the number of actions and T the number of time-steps. This result improves over the best previous known bound Õ(HS √(AT)) achieved by the UCRL2 algorithm of Jaksch et al., 2010. The key significance of our new results is that when T≥ H^3S^3A and SA≥ H, it leads to a regret of Õ(√(HSAT)) that matches the established lower bound of Ω(√(HSAT)) up to a logarithmic factor. Our analysis contains two key insights. We use careful application of concentration inequalities to the optimal value function as a whole, rather than to the transitions probabilities (to improve scaling in S), and we define Bernstein-based "exploration bonuses" that use the empirical variance of the estimated values at the next states (to improve scaling in H).

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset