Model-based clustering for populations of networks
We propose a model-based clustering method for populations of networks that describes the joint distribution of a sequence of networks in a parsimonious manner, and can be used to identify subpopulations of networks that share certain topological properties of interest. We discuss how maximum likelihood estimation can be performed with the EM algorithm and study the performance of the proposed method on simulated data. We conclude with an example application to a sequence of face-to-face interaction networks measured in an office environment.
READ FULL TEXT