Multi-Agent Determinantal Q-Learning

Yaodong Yang, Ying Wen, Jun Wang, Liheng Chen, Kun Shao, David Mguni, Weinan Zhang
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:10757-10766, 2020.

Abstract

Centralized training with decentralized execution has become an important paradigm in multi-agent learning. Though practical, current methods rely on restrictive assumptions to decompose the centralized value function across agents for execution. In this paper, we eliminate this restriction by proposing multi-agent determinantal Q-learning. Our method is established on Q-DPP, a novel extension of determinantal point process (DPP) to multi-agent setting. Q-DPP promotes agents to acquire diverse behavioral models; this allows a natural factorization of the joint Q-functions with no need for \emph{a priori} structural constraints on the value function or special network architectures. We demonstrate that Q-DPP generalizes major solutions including VDN, QMIX, and QTRAN on decentralizable cooperative tasks. To efficiently draw samples from Q-DPP, we develop a linear-time sampler with theoretical approximation guarantee. Our sampler also benefits exploration by coordinating agents to cover orthogonal directions in the state space during training. We evaluate our algorithm on multiple cooperative benchmarks; its effectiveness has been demonstrated when compared with the state-of-the-art.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-yang20i, title = {Multi-Agent Determinantal Q-Learning}, author = {Yang, Yaodong and Wen, Ying and Wang, Jun and Chen, Liheng and Shao, Kun and Mguni, David and Zhang, Weinan}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {10757--10766}, 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/yang20i/yang20i.pdf}, url = {http://proceedings.mlr.press/v119/yang20i.html}, abstract = {Centralized training with decentralized execution has become an important paradigm in multi-agent learning. Though practical, current methods rely on restrictive assumptions to decompose the centralized value function across agents for execution. In this paper, we eliminate this restriction by proposing multi-agent determinantal Q-learning. Our method is established on Q-DPP, a novel extension of determinantal point process (DPP) to multi-agent setting. Q-DPP promotes agents to acquire diverse behavioral models; this allows a natural factorization of the joint Q-functions with no need for \emph{a priori} structural constraints on the value function or special network architectures. We demonstrate that Q-DPP generalizes major solutions including VDN, QMIX, and QTRAN on decentralizable cooperative tasks. To efficiently draw samples from Q-DPP, we develop a linear-time sampler with theoretical approximation guarantee. Our sampler also benefits exploration by coordinating agents to cover orthogonal directions in the state space during training. We evaluate our algorithm on multiple cooperative benchmarks; its effectiveness has been demonstrated when compared with the state-of-the-art.} }
Endnote
%0 Conference Paper %T Multi-Agent Determinantal Q-Learning %A Yaodong Yang %A Ying Wen %A Jun Wang %A Liheng Chen %A Kun Shao %A David Mguni %A Weinan Zhang %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-yang20i %I PMLR %P 10757--10766 %U http://proceedings.mlr.press/v119/yang20i.html %V 119 %X Centralized training with decentralized execution has become an important paradigm in multi-agent learning. Though practical, current methods rely on restrictive assumptions to decompose the centralized value function across agents for execution. In this paper, we eliminate this restriction by proposing multi-agent determinantal Q-learning. Our method is established on Q-DPP, a novel extension of determinantal point process (DPP) to multi-agent setting. Q-DPP promotes agents to acquire diverse behavioral models; this allows a natural factorization of the joint Q-functions with no need for \emph{a priori} structural constraints on the value function or special network architectures. We demonstrate that Q-DPP generalizes major solutions including VDN, QMIX, and QTRAN on decentralizable cooperative tasks. To efficiently draw samples from Q-DPP, we develop a linear-time sampler with theoretical approximation guarantee. Our sampler also benefits exploration by coordinating agents to cover orthogonal directions in the state space during training. We evaluate our algorithm on multiple cooperative benchmarks; its effectiveness has been demonstrated when compared with the state-of-the-art.
APA
Yang, Y., Wen, Y., Wang, J., Chen, L., Shao, K., Mguni, D. & Zhang, W.. (2020). Multi-Agent Determinantal Q-Learning. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:10757-10766 Available from http://proceedings.mlr.press/v119/yang20i.html.

Related Material