Model-free Reinforcement Learning in Infinite-horizon Average-reward Markov Decision Processes
This work addresses efficient learning in large-scale reinforcement learning problems, offering improved regret bounds for model-free methods, though it is incremental with specific algorithmic advancements.
The paper tackles the problem of model-free reinforcement learning in infinite-horizon average-reward Markov Decision Processes by introducing two algorithms: one achieves O(T^{2/3}) regret under minimal assumptions, and the other improves to O(sqrt{T}) regret under stronger ergodic assumptions, significantly outperforming the prior O(T^{3/4}) regret.
Model-free reinforcement learning is known to be memory and computation efficient and more amendable to large scale problems. In this paper, two model-free algorithms are introduced for learning infinite-horizon average-reward Markov Decision Processes (MDPs). The first algorithm reduces the problem to the discounted-reward version and achieves $\mathcal{O}(T^{2/3})$ regret after $T$ steps, under the minimal assumption of weakly communicating MDPs. To our knowledge, this is the first model-free algorithm for general MDPs in this setting. The second algorithm makes use of recent advances in adaptive algorithms for adversarial multi-armed bandits and improves the regret to $\mathcal{O}(\sqrt{T})$, albeit with a stronger ergodic assumption. This result significantly improves over the $\mathcal{O}(T^{3/4})$ regret achieved by the only existing model-free algorithm by Abbasi-Yadkori et al. (2019a) for ergodic MDPs in the infinite-horizon average-reward setting.