Bayesian model-based clustering for multiple network data

There is increasing appetite for analysing multiple network data. This is due to the fast-growing body of applications that demand such methods. These include: the study of connectomes in neuroscience and the study of human mobility with respect to intelligent displays in computer science. Recent technological advancements have allowed the collection of this type of data. Both applications entail the analysis of a heterogeneous population of networks. In this paper we focus on the problem of clustering the elements of a network population, here each cluster will be characterised by a network representative. We take advantage of the Bayesian machinery to simultaneously infer the cluster membership, the representatives and the community structure of the representatives. Extensive simulation studies show our model performs well in both clustering multiple network data and inferring the model parameters. , we of false positive , 0.2, different to its true while its 95% credible interval covers a wide range of values, indicating that our MCMC chain struggles to make inferences here. These results suggest that our model performs well in most cases, even for high noise levels, but we must be cautious when making inferences for network populations with great variability in their structure. of initialise the nodes’ block membership of the representatives using R the presence of two underlying blocks. here that a simpler network model, namely the Erd¨os-R´enyi model, could also be applied to describe the representative networks. MCMC 500,000 iterations a burn-in