a social learning model for multi- hypothesis decision making

This paper deals with the problem of social learning and distributed estimation. a network of agents having limited cognitive abilities try to find the true hypothesis based on their noisy private observations and their social observations. studying the formation and evolution of beliefs in social networks is an inherently interdisciplinary field. in most of the learning models, agents directly communicate to know each other's beliefs, but we assume, agents observe each other's decisions and have to infer other agents' belief. agents use simplified-bayesian learning to infer information from neighbors' decisions to learn the true hypothesis. the model is developed for multi-hypothesis scenarios, has a reasonable computational complexity, and reduces the herd behavior caused by the cognitive characteristics of the agents. monte carlo simulation results show that by using the model, the agents learn the truth and reach a belief consensus.