Clustering Networks with Node and Edge Attributes using Bregman Divergence

Abstract

Network clustering tackles the problem of identifying sets of nodes (clusters or communities) that have similar connection patterns. However, in many modern scenarios, nodes also have attributes that are correlated with the network structure. Thus, network information (edges) and node information (attributes) can be jointly leveraged to design high-performance clustering algorithms. Under a general model for the network and node attributes, this thesis establishes an information-theoretic criterion for the exact recovery of community labels and characterizes a phase transition determined by the Chernoff-Hellinger divergence of the model. The criterion shows how network and attribute information can be exchanged in order to yield exact recovery (e.g., more reliable network information requires less reliable attribute information). This thesis also presents two iterative clustering algorithms that greedily maximizes the joint likelihood of the model under the assumption that the probability distribution of network edges and node attributes belong to exponential families. Extensive analysis of the two algorithms on both synthetic datasets and real benchmarks highlights their accuracy and performance with respect to other state-of-the-art approaches.