Social Learning in Multi Agent Multi Armed Bandits

We introduce a novel decentralized, multi agent version of the classical Multi-Arm Bandit (MAB) problem, consisting of n agents, that collaboratively and simultaneously solve the same instance of K armed MAB to minimize individual regret. The agents can communicate and collaborate among each other only through a pairwise asynchronous gossip based protocol that exchange a limited number of bits. In our model, agents at each point decide on (i) which arm to play, (ii) whether to, and if so (iii) what and whom to communicate with. We develop a novel algorithm in which agents, whenever they choose, communicate only arm-ids and not samples, with another agent chosen uniformly and independently at random. The peragent regret achieved by our algorithm is O(⌈K/n⌉ + log(n)/Δ log(T)), where Δ is the difference between the mean of the best and second best arm. Furthermore, any agent in our algorithm communicates (arm-ids to an uniformly and independently chosen agent) only a total of Θ(log(T)) times over a time interval of T. We compare our results to two benchmarks - one where there is no communication among agents and one corresponding to complete interaction, where an agent has access to the entire system history of arms played and rewards obtained of all agents. We show both theoretically and empirically, that our algorithm experiences a significant reduction both in per-agent regret when compared to the case when agents do not collaborate and each agent is playing the standard MAB problem (where regret would scale linearly in K), and in communication complexity when compared to the full interaction setting which requires T communication attempts by an agent over T arm pulls. Our result thus demonstrates that even a minimal level of collaboration among the different agents enables a significant reduction in per-agent regret.