Time-varying diffusion social learning

Abstract

This paper studies a time-varying diffusion social learning algorithm, for an agent in a network trying to learn a true hypothesis (or state). We will show that agents in a weakly connected sub-network will be under the influence of agents in the strongly connected sub-network, which leads to a mind control scenario. We will derive analytical expressions of the limiting beliefs of the agents in the time-varying network. Moreover, we will extend the results of time-varying weakly connected networks to the case where agents assign decreasing weights to neighboring agents.