Multiagent Soft Q-Learning

Ermo Wei, Drew Wicke, David Freelan +1

Policy gradient methods are often applied to reinforcement learning in continuous multiagent games. These methods perform local search in the joint-action space, and as we show, th…