A guided tour of Chernoff bounds
Information Processing Letters
An Upper Bound on the Loss from Approximate Optimal-Value Functions
Machine Learning
Finite-sample convergence rates for Q-learning and indirect algorithms
Proceedings of the 1998 conference on Advances in neural information processing systems II
Markov Decision Processes: Discrete Stochastic Dynamic Programming
Markov Decision Processes: Discrete Stochastic Dynamic Programming
Introduction to Reinforcement Learning
Introduction to Reinforcement Learning
Neuro-Dynamic Programming
The Sample Complexity of Exploration in the Multi-Armed Bandit Problem
The Journal of Machine Learning Research
Prediction, Learning, and Games
Prediction, Learning, and Games
The Journal of Machine Learning Research
Dynamic Programming and Optimal Control, Vol. II
Dynamic Programming and Optimal Control, Vol. II
Reinforcement Learning in Finite MDPs: PAC Analysis
The Journal of Machine Learning Research
REGAL: a regularization based algorithm for reinforcement learning in weakly communicating MDPs
UAI '09 Proceedings of the Twenty-Fifth Conference on Uncertainty in Artificial Intelligence
Algorithms for Reinforcement Learning
Algorithms for Reinforcement Learning
Near-optimal Regret Bounds for Reinforcement Learning
The Journal of Machine Learning Research
PAC bounds for discounted MDPs
ALT'12 Proceedings of the 23rd international conference on Algorithmic Learning Theory
Hi-index | 0.00 |
We consider the problems of learning the optimal action-value function and the optimal policy in discounted-reward Markov decision processes (MDPs). We prove new PAC bounds on the sample-complexity of two well-known model-based reinforcement learning (RL) algorithms in the presence of a generative model of the MDP: value iteration and policy iteration. The first result indicates that for an MDP with N state-action pairs and the discount factor 驴驴[0,1) only O(Nlog(N/驴)/((1驴驴)3 驴 2)) state-transition samples are required to find an 驴-optimal estimation of the action-value function with the probability (w.p.) 1驴驴. Further, we prove that, for small values of 驴, an order of O(Nlog(N/驴)/((1驴驴)3 驴 2)) samples is required to find an 驴-optimal policy w.p. 1驴驴. We also prove a matching lower bound of 驴(Nlog(N/驴)/((1驴驴)3 驴 2)) on the sample complexity of estimating the optimal action-value function with 驴 accuracy. To the best of our knowledge, this is the first minimax result on the sample complexity of RL: the upper bounds match the lower bound in terms of N, 驴, 驴 and 1/(1驴驴) up to a constant factor. Also, both our lower bound and upper bound improve on the state-of-the-art in terms of their dependence on 1/(1驴驴).