Bayesian block modelling for weighted networks
Ian Gallagher · 2010
This paper presents a Bayesian approach to block modelling weighted networks to identify role assignments. This data arises commonly in many forms of social networks where we have a count of the number of communications between users. By using Variational Bayes techniques, we are able to perform fast approximate posterior inference that allows us to recover the underlying role groups in the network and their interactions. We apply our method to synthetic and real communication networks, in particular the Enron email data set.