@inproceedings{249bd1b4f2814b628891d1b1daaa1620,
title = "PAC bounds for discounted MDPs",
abstract = "We study upper and lower bounds on the sample-complexity of learning near-optimal behaviour in finite-state discounted Markov Decision Processes (mdps). We prove a new bound for a modified version of Upper Confidence Reinforcement Learning (ucrl) with only cubic dependence on the horizon. The bound is unimprovable in all parameters except the size of the state/action space, where it depends linearly on the number of non-zero transition probabilities. The lower bound strengthens previous work by being both more general (it applies to all policies) and tighter. The upper and lower bounds match up to logarithmic factors provided the transition matrix is not too dense.",
keywords = "Markov decision processes, PAC-MDP, Reinforcement learning, exploration exploitation, sample-complexity",
author = "Tor Lattimore and Marcus Hutter",
year = "2012",
doi = "10.1007/978-3-642-34106-9\_26",
language = "English",
isbn = "9783642341052",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "320--334",
booktitle = "Algorithmic Learning Theory - 23rd International Conference, ALT 2012, Proceedings",
note = "23rd International Conference on Algorithmic Learning Theory, ALT 2012 ; Conference date: 29-10-2012 Through 31-10-2012",
}