English

Refined Analysis of FPL for Adversarial Markov Decision Processes

Machine Learning 2020-08-24 v1 Machine Learning

Abstract

We consider the adversarial Markov Decision Process (MDP) problem, where the rewards for the MDP can be adversarially chosen, and the transition function can be either known or unknown. In both settings, Follow-the-PerturbedLeader (FPL) based algorithms have been proposed in previous literature. However, the established regret bounds for FPL based algorithms are worse than algorithms based on mirrordescent. We improve the analysis of FPL based algorithms in both settings, matching the current best regret bounds using faster and simpler algorithms.

Keywords

Cite

@article{arxiv.2008.09251,
  title  = {Refined Analysis of FPL for Adversarial Markov Decision Processes},
  author = {Yuanhao Wang and Kefan Dong},
  journal= {arXiv preprint arXiv:2008.09251},
  year   = {2020}
}

Comments

11 pages