Information Sharing and Searching via Collaborative Reinforcement Learning

摘要

This paper proposes a method for computing a routing policy-value function for effective information sharing and searching in arbitrary networks of agents through collaborative reinforcement learning. This is done by means of local computations performed by agents and payoff propagation. The aim is to 'tune' a network of agents for efficient and effective information searching and sharing, without altering the topology or imposing an overlay structure.