Fusion and Clustering of Multilayer Graphs through Regularized Bistochastic Approximation (Under Review)

Multilayer graphs encode multiple different types of relationships and interactions between a set of entities. In this paper, we propose a novel principled method of regularized bistochastic approximation in order to fuse multilayer graphs and divide this unified representation into clusters. Our proposed method learns an approximated graph from multiple layers through bistochastic approximation and identifies underlying block structures in the learned graph through regularization for graph clustering in the unified optimization formulation. We also implement a new iterative algorithm to solve this formulated problem, which is guaranteed to converge to the optimal solution. We evaluate our approach on six public benchmark datasets for multilayer graph clustering. The experimental results show that our approach obtains state-of-the-art clustering performance and is robust to noise in the multilayer graphs.

Citation: Fusion and Clustering of Multilayer Graphs through Regularized Bistochastic Approximation. Brian Reily and Hao Zhang. Information Fusion, Submitted 2021.