Planning in Markov Decision Processes with Gap-Dependent Sample Complexity

Abstract

We propose MDP-GapE, a new trajectory-based Monte-Carlo Tree Search algorithm for planning in a Markov Decision Process in which transitions have a finite support. We prove an upper bound on the number of sampled trajectories needed for MDP-GapE to identify a near-optimal action with high probability. This problem-dependent result is expressed in terms of the sub-optimality gaps of the state-action pairs that are visited during exploration. Our experiments reveal that MDP-GapE is also effective in practice, in contrast with other algorithms with sample complexity guarantees in the fixed-confidence setting, that are mostly theoretical.

Cite

Text

Jonsson et al. "Planning in Markov Decision Processes with Gap-Dependent Sample Complexity." Neural Information Processing Systems, 2020.

Markdown

[Jonsson et al. "Planning in Markov Decision Processes with Gap-Dependent Sample Complexity." Neural Information Processing Systems, 2020.](https://mlanthology.org/neurips/2020/jonsson2020neurips-planning/)

BibTeX

@inproceedings{jonsson2020neurips-planning,
  title     = {{Planning in Markov Decision Processes with Gap-Dependent Sample Complexity}},
  author    = {Jonsson, Anders and Kaufmann, Emilie and Menard, Pierre and Domingues, Omar Darwiche and Leurent, Edouard and Valko, Michal},
  booktitle = {Neural Information Processing Systems},
  year      = {2020},
  url       = {https://mlanthology.org/neurips/2020/jonsson2020neurips-planning/}
}