An anytime algorithm for trust propagation in social networks

Abstract

Trust propagation in social networks is a challenging task. It is difficult to model human trust, and the data is huge and very sparse. Due to these challenges, the algorithms available to propagate trust have complexity issues. We used the MRFTrust algorithm created by Tosun and Sheppard to produce an anytime algorithm for trust propagation. To do this we use sampling techniques and increased horizon size to reduce the complexity and decrease runtimes. We show that we can dramatically reduce the number of nodes considered in the algorithm, yet still achieve a superior result.