Learning on dynamic social network

In this paper, we study the formation and update process of agent´s opinion of view in the social network. Based on the Non-Bayesian learning model, we develop three kinds of social learning models: Static social network model with fixed delay; Random evolution of dynamic social network model; Deterministic evolution of dynamic social network model with fixed delay. By applying the algebraic graph theory, the non-negative matrix theory and the probability theory, we analytically prove that as long as individuals set their final beliefs to be a linear combination of the Bayesian posterior beliefs and the opinions of his neighbors, they can aggregate information successfully and learn the true state of the world.