Academic Report of School of Mathematical Sciences [2026] No. 030

(Series Report for High-Level University Construction No. 1289)

Title：Methods and Theories for Community Detection in Network Data

Speaker：Ying Yang, Professor ( Tsinghua University)

Time：14:30-15:30, Apr. 16, 2026

Location：Tencent Meeting ID 707293202

Abstract：Over the past decade, research on network data has yielded significant theoretical and applied results in fields such as the social sciences, biology, genetics, and statistics. From social networks depicting interpersonal relationships to biological networks describing protein interactions, network data is virtually ubiquitous. Uncovering the intrinsic structure of network data and performing community detection holds important theoretical and practical value. This presentation focuses on community detection problems in three types of network data. The first category involves community detection in weighted networks. We make nonparametric assumptions regarding the weight distribution in the weighted random block model. Within this framework, we design community detection methods for scenarios where the number of communities is known or unknown, and theoretically prove that the proposed methods possess relationship-consistency. The second category concerns outlier detection in network data. For network data with imbalanced community structures, we define small-scale communities within the random clustering framework and propose a small-scale detection method based on node bias; this method not only accurately identifies small-scale communities but also effectively estimates the community structure within larger clusters. The third category is community detection in multilayer network data. For multilayer directed networks, we propose a collaborative spectral clustering algorithm capable of capturing the asymmetric community structure in directed networks, and we prove the consistency of community estimates under a multilayer stochastic co-blocking model. Furthermore, to address the analytical demands of large-scale multilayer networks, we further propose a stochastic spectral clustering algorithm that can efficiently perform community detection in multilayer networks with millions of nodes.

Speaker Profile：Ying Yang is a Professor of Statistics and Data Science at Tsinghua University. Her research interests include nonparametric and semiparametric estimation, smooth parameter selection, median regression, longitudinal data analysis, and network data analysis. She has served as Vice Chair of the 11th and 12th Councils of the Chinese Mathematical Society’s Probability and Statistics Branch, and as a member of the 2nd and 3rd National Steering Committees for Graduate Education in Applied Statistics.

Faculty and students are welcome to attend!

Invited by: Yan Zhou

School of Mathematical Sciences

April 13, 2026