B1018
Title: Community detection for directed networks with awareness of node popularity
Authors: Ting Li - Hong Kong Polytechnic University (Hong Kong) [presenting]
Abstract: Community detection for directed networks has raised lots interests in recent decades. While many of the existing results introduced block-wise structure, most of them have the restriction that nodes in the same community to behave identically or change uniformly in all communities. However, the heterogeneous node popularity is widely observed both in undirected and directed networks, and has been studied under undirected scenarios. Motivated by the variability of node popularity in empirical directed networks, we propose a novel probabilistic framework for directed network community detection, called the two-way node popularity model (TNPM). To fit the proposed model, we develop the Rank One Approximation Algorithm (ROA) and establish the consistency of ROA for community detection. In addition, an alternative computationally efficient algorithm, called Two-Stage Divided Cosine Algorithm (TSDC), is proposed to fit large-scale networks. Extensive numerical studies demonstrate the advantages of our proposed method in terms of both estimation accuracy and computation efficiency. The method is also applied to the Worldwide Trading Networks, yielding some interesting findings.