In the paper Bayesian Action Decoder for Deep Multi-Agent Reinforcement Learning, it is written

We allow centralised training but require decentralised execution, from which follows that the policies $\pi^a$ are known to all agents.

My confusion stems from e.g. the following scenario: In an $N$ player game, all $N$ players share parameters in a single agent network. In such a scenario, when we move on to decentralised execution, how does this take place if parameters for all $N$ players were shared across a single network during training?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.