On Thompson sampling and asymptotic optimality

Date

2017

Authors

Leike, Jan
Lattimore, Tor
Orseau, Laurent
Hutter, Marcus

Journal Title

Journal ISSN

Volume Title

Publisher

International Joint Conferences on Artificial Intelligence

Abstract

We discuss some recent results on Thompson sampling for nonparametric reinforcement learning in countable classes of general stochastic environments. These environments can be non-Markovian, non-ergodic, and partially observable. We show that Thompson sampling learns the environment class in the sense that (1) asymptotically its value converges in mean to the optimal value and (2) given a recoverability assumption regret is sublinear. We conclude with a discussion about optimality in reinforcement learning.

Description

Keywords

Citation

Source

Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)

Type

Conference paper

Book Title

Entity type

Access Statement

Free Access via Publisher Site

License Rights

Restricted until

2099-12-31

Downloads

File
Description