Modeling Network Populations via Graph Distances

摘要: This article introduces a new class of models for multiple networks. The core idea is to parametrize distribution on labelled graphs in terms Frechet mean graph (which depends user-specified choice metric or distance) and parameter that controls the concentration this about its mean. Entropy natural such control, varying from point mass concentrated itself uniform over all given vertex set. We provide hierarchical Bayesian approach exploiting construction, along with straightforward strategies sampling resultant posterior distribution. conclude by demonstrating efficacy our via simulation studies two multiple-network data analysis examples: one drawn systems biology other neuroscience.