Non-Bayesian Learning in Social Networks with Time-varying Weights

This paper investigates an social learning model with time-varying weights, in which the individual updates her belief through observing private signal caused by social event and communicating with those regarded as neighbors in the sense of network topology. The private signal is involved in the updating law through Bayes' rule. During the communication with neighbors, the individual obtains weighted average of others' beliefs. Using the convergence property of the transition matrix and coefficient of ergodicity, we show that, under mild assumptions, repeated observation and communications can lead beliefs of the entire group to the true state of the social event.