Efficiently Solving MDPs with Stochastic Mirror Descent
Abstract
We present a unified framework based on primal-dual stochastic mirror descent for approximately solving infinite-horizon Markov decision processes (MDPs) given a generative model. When applied to an average-reward MDP with $A_{tot}$ total actions and mixing time bound $t_{mix}$ our method computes an $\epsilon$-optimal policy with an expected $\widetilde{O}(t_{mix}^2 A_{tot} \epsilon^{-2})$ samples from the state-transition matrix, removing the ergodicity dependence of prior art. When applied to a $\gamma$-discounted MDP with $A_{tot}$ total actions our method computes an $\epsilon$-optimal policy with an expected $\widetilde{O}((1-\gamma)^{-4} A_{tot} \epsilon^{-2})$ samples, improving over the best-known primal-dual methods while matching the state-of-the-art up to a $(1-\gamma)^{-1}$ factor. Both methods are model-free, update state values and policies simultaneously, and run in time linear in the number of samples taken. We achieve these results through a more general stochastic mirror descent framework for solving bilinear saddle-point problems with simplex and box domains and we demonstrate the flexibility of this framework by providing further applications to constrained MDPs.
BibTeX
@InProceedings{pmlr-v119-jin20f,
title = {Efficiently Solving {MDP}s with Stochastic Mirror Descent},
author = {Jin, Yujia and Sidford, Aaron},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
pages = {4890--4900},
year = {2020},
editor = {III, Hal Daumé and Singh, Aarti},
volume = {119},
series = {Proceedings of Machine Learning Research},
month = {13--18 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v119/jin20f/jin20f.pdf},
url = {https://proceedings.mlr.press/v119/jin20f.html},
abstract = {We present a unified framework based on primal-dual stochastic mirror descent for approximately solving infinite-horizon Markov decision processes (MDPs) given a generative model. When applied to an average-reward MDP with $A_{tot}$ total actions and mixing time bound $t_{mix}$ our method computes an $\epsilon$-optimal policy with an expected $\widetilde{O}(t_{mix}^2 A_{tot} \epsilon^{-2})$ samples from the state-transition matrix, removing the ergodicity dependence of prior art. When applied to a $\gamma$-discounted MDP with $A_{tot}$ total actions our method computes an $\epsilon$-optimal policy with an expected $\widetilde{O}((1-\gamma)^{-4} A_{tot} \epsilon^{-2})$ samples, improving over the best-known primal-dual methods while matching the state-of-the-art up to a $(1-\gamma)^{-1}$ factor. Both methods are model-free, update state values and policies simultaneously, and run in time linear in the number of samples taken. We achieve these results through a more general stochastic mirror descent framework for solving bilinear saddle-point problems with simplex and box domains and we demonstrate the flexibility of this framework by providing further applications to constrained MDPs.}
}