Academy of Mathematics and Systems Science, CAS
Colloquia & Seminars

Speaker:	李挺博士,香港理工大学
Inviter:	潘文亮
Title:	Two-way Node Popularity Model for Directed and Bipartite Networks
Language:	Chinese
Time & Venue:	2023.07.13 16:00-17:00 思源楼S625
Abstract:	Community detection for directed and bipartite networks has received considerable attention in recent decades. However, most of the existing studies overlook the heterogeneity of an important network feature, node popularity, and implicitly restrict that nodes within the same community behave identically or change uniformly across all communities. To account for the variability of node popularity in empirical directed and bipartite networks, we propose a novel probabilistic framework for these two types of data, called the two-way node popularity model (TNPM). For model fitting and community structure uncovering, we develop the Rank-one Approximation Algorithm (ROA) and establish its consistency for probability matrix estimation and community detection. In addition, we introduce an alternative more computationally efficient algorithm, the Two-Stage Divided Cosine Algorithm (TSDC), to handle the large-scale networks. Extensive numerical studies demonstrate the advantages of our proposed methods in terms of both estimation accuracy and computation efficiency. Finally, we apply our approach to the Worldwide Trading Networks and MovieLens 100K Dataset, and obtain interesting findings.