arXiv (Cornell University) · 2019 · 43 citations · 37 references
ComputingAutonomous SystemsComputer ScienceArtificial IntelligenceGame TheoryMachine LearningCoordination GamesDeep LearningMulti-agent Robot LearningMulti-agent SystemsReinforcement LearningReinforcement Learning FoundationsCoordination GraphsDeep Reinforcement LearningLow-rank ApproximationMulti-agent LearningMulti-agent Reinforcement LearningRepresentational CapacitySample EfficiencyValue Function Generalization
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.
37
Human-level control through deep reinforcement learning
Volodymyr Mnih, Koray Kavukcuoglu, David Silver et al. · Nature · 2015 · 30K citations
Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
Jun‐Young Chung, Çağlar Gülçehre, Kyunghyun Cho et al. · arXiv (Cornell University) · 2014 · 10.8K citations · Full text
Grandmaster level in StarCraft II using multi-agent reinforcement learning
Oriol Vinyals, I. Babuschkin, Wojciech Marian Czarnecki et al. · Nature · 2019 · 3.4K citations