Conference item
Deep coordination graphs
- Abstract:
- This paper introduces the deep coordination graph (DCG) for collaborative multi-agent reinforcement learning. DCG strikes a flexible trade-off between representational capacity and generalization by factoring the joint value function of all agents according to a coordination graph into payoffs between pairs of agents. The value can be maximized by local message passing along the graph, which allows training of the value function end-to-end with Q-learning. Payoff functions are approximated with deep neural networks that employ parameter sharing and low-rank approximations to significantly improve sample efficiency. We show that DCG can solve predator-prey tasks that highlight the relative overgeneralization pathology, as well as challenging StarCraft II micromanagement tasks.
- Publication status:
- Published
- Peer review status:
- Peer reviewed
Actions
Access Document
- Files:
-
-
(Preview, Accepted manuscript, 960.9KB, Terms of use)
-
- Publication website:
- http://proceedings.mlr.press/v119/boehmer20a.html
Authors
- Publisher:
- Journal of Machine Learning Research
- Host title:
- International Conference on Machine Learning, 13-18 July 2020, Virtual
- Pages:
- 980-991
- Series:
- Proceedings of Machine Learning Research
- Series number:
- 119
- Publication date:
- 2020-11-21
- Acceptance date:
- 2020-06-01
- Event title:
- 37th International Conference on Machine Learning (ICML 2020)
- Event location:
- Virtual
- Event website:
- https://icml.cc/Conferences/2020
- Event start date:
- 2020-07-12
- Event end date:
- 2020-07-18
- ISSN:
-
2640-3498
- Language:
-
English
- Keywords:
- Pubs id:
-
1118781
- Local pid:
-
pubs:1118781
- Deposit date:
-
2020-07-15
Terms of use
- Copyright holder:
- Boehmer, w et al.
- Copyright date:
- 2020
- Rights statement:
- © 2020 The Authors.
- Notes:
- This paper was presented at the 37th International Conference on Machine Learning (ICML 2020), 12-18 July 2020. This is the accepted manuscript version of the paper. The final version is available online from PMLR at: http://proceedings.mlr.press/v119/boehmer20a.html
If you are the owner of this record, you can report an update to it here: Report update to this record