Using Active Queries to Learn Local Stochastic Behaviors in Social Networks.

Using a stochastic synchronous dynamical system (SyDS) as a formal model, we study the problem of inferring local behaviors of nodes in networked social systems. We focus on probabilistic threshold functions as local functions. We use an active query mechanism where a user interacts with the system by submitting queries. We develop an efficient algorithm that infers the probabilistic threshold functions using the responses to the queries. Our algorithm generates provably good query sets. We also present experimental results to demonstrate the performance of our algorithm.