A minimax and asymptotically optimal algorithm for stochastic bandits
Machine Learning
2017-09-21 v2 Machine Learning
Statistics Theory
Statistics Theory
Abstract
We propose the kl-UCB ++ algorithm for regret minimization in stochastic bandit models with exponential families of distributions. We prove that it is simultaneously asymptotically optimal (in the sense of Lai and Robbins' lower bound) and minimax optimal. This is the first algorithm proved to enjoy these two properties at the same time. This work thus merges two different lines of research with simple and clear proofs.
Cite
@article{arxiv.1702.07211,
title = {A minimax and asymptotically optimal algorithm for stochastic bandits},
author = {Pierre Ménard and Aurélien Garivier},
journal= {arXiv preprint arXiv:1702.07211},
year = {2017}
}