Off-Policy Actor-Critic with Shared Experience Replay

Simon Schmitt, Matteo Hessel, Karen Simonyan
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:8545-8554, 2020.

Abstract

We investigate the combination of actor-critic reinforcement learning algorithms with a uniform large-scale experience replay and propose solutions for two ensuing challenges: (a) efficient actor-critic learning with experience replay (b) the stability of off-policy learning where agents learn from other agents behaviour. To this end we analyze the bias-variance tradeoffs in V-trace, a form of importance sampling for actor-critic methods. Based on our analysis, we then argue for mixing experience sampled from replay with on-policy experience, and propose a new trust region scheme that scales effectively to data distributions where V-trace becomes unstable. We provide extensive empirical validation of the proposed solutions on DMLab-30 and further show the benefits of this setup in two training regimes for Atari: (1) a single agent is trained up until 200M environment frames per game (2) a population of agents is trained up until 200M environment frames each and may share experience. We demonstrate state-of-the-art data efficiency among model-free agents in both regimes.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-schmitt20a, title = {Off-Policy Actor-Critic with Shared Experience Replay}, author = {Schmitt, Simon and Hessel, Matteo and Simonyan, Karen}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {8545--8554}, 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/schmitt20a/schmitt20a.pdf}, url = {https://proceedings.mlr.press/v119/schmitt20a.html}, abstract = {We investigate the combination of actor-critic reinforcement learning algorithms with a uniform large-scale experience replay and propose solutions for two ensuing challenges: (a) efficient actor-critic learning with experience replay (b) the stability of off-policy learning where agents learn from other agents behaviour. To this end we analyze the bias-variance tradeoffs in V-trace, a form of importance sampling for actor-critic methods. Based on our analysis, we then argue for mixing experience sampled from replay with on-policy experience, and propose a new trust region scheme that scales effectively to data distributions where V-trace becomes unstable. We provide extensive empirical validation of the proposed solutions on DMLab-30 and further show the benefits of this setup in two training regimes for Atari: (1) a single agent is trained up until 200M environment frames per game (2) a population of agents is trained up until 200M environment frames each and may share experience. We demonstrate state-of-the-art data efficiency among model-free agents in both regimes.} }
Endnote
%0 Conference Paper %T Off-Policy Actor-Critic with Shared Experience Replay %A Simon Schmitt %A Matteo Hessel %A Karen Simonyan %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-schmitt20a %I PMLR %P 8545--8554 %U https://proceedings.mlr.press/v119/schmitt20a.html %V 119 %X We investigate the combination of actor-critic reinforcement learning algorithms with a uniform large-scale experience replay and propose solutions for two ensuing challenges: (a) efficient actor-critic learning with experience replay (b) the stability of off-policy learning where agents learn from other agents behaviour. To this end we analyze the bias-variance tradeoffs in V-trace, a form of importance sampling for actor-critic methods. Based on our analysis, we then argue for mixing experience sampled from replay with on-policy experience, and propose a new trust region scheme that scales effectively to data distributions where V-trace becomes unstable. We provide extensive empirical validation of the proposed solutions on DMLab-30 and further show the benefits of this setup in two training regimes for Atari: (1) a single agent is trained up until 200M environment frames per game (2) a population of agents is trained up until 200M environment frames each and may share experience. We demonstrate state-of-the-art data efficiency among model-free agents in both regimes.
APA
Schmitt, S., Hessel, M. & Simonyan, K.. (2020). Off-Policy Actor-Critic with Shared Experience Replay. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:8545-8554 Available from https://proceedings.mlr.press/v119/schmitt20a.html.

Related Material