We propose a generalized stochastic block model to explore the mesoscopicstructures in signed networks by grouping vertices that exhibit similarpositive and negative connection profiles into the same cluster. In this model,the group memberships are viewed as hidden or unobserved quantities, and theconnection patterns between groups are explicitly characterized by two blockmatrices, one for positive links and the other for negative links. By fittingthe model to the observed network, we can not only extract various structuralpatterns existing in the network without prior knowledge, but also recognizewhat specific structures we obtained. Furthermore, the model parameters providevital clues about the probabilities that each vertex belongs to differentgroups and the centrality of each vertex in its corresponding group. Thisinformation sheds light on the discovery of the networks' overlappingstructures and the identification of two types of important vertices, whichserve as the cores of each group and the bridges between different groups,respectively. Experiments on a series of synthetic and real-life networks showthe effectiveness as well as the superiority of our model.
展开▼