Skip to search formSkip to main contentSkip to account menu

Achieving Tractable Minimax Optimal Regret in Average Reward MDPs

This paper presents the first tractable algorithm with minimax optimal regret of $\widetilde{\mathrm{O}}(\sqrt{\mathrm{sp}(h^*) S A T})$, and relies on a novel subroutine, Projected Mitigated Extended Value Iteration (PMEVI), to compute bias-constrained optimal policies efficiently.
arXiv (opens in a new tab)

Logarithmic Regret of Exploration in Average Reward Markov Decision Processes

The management of bad episodes is much better under (VM) than (DT) by making the regret of exploration logarithmic rather than linear, made possible by a new in-depth understanding of the contrasting behaviors of confidence regions during good and bad episodes.
arXiv (opens in a new tab)

The regret lower bound for communicating Markov Decision Processes

This paper proves that the regret lower bound becomes significatively more complex in communicating MDPs, and revisits the necessary explorative behavior of consistent learning agents and explains that all optimal regions of the environment must be overvisited compared to sub-optimal ones.
arXiv (opens in a new tab)
By clicking accept or continuing to use the site, you agree to the terms outlined in our Privacy Policy (opens in a new tab), Terms of Service (opens in a new tab), and Dataset License (opens in a new tab)