Year: 2025
Type: conference-paper
Abstract: Identifying community structures is vital in analysing and understanding the functions of complex networks. Topological information is widely used in most existing community detection algorithms; however, it is important to simultaneously consider the available node attribute information to capture the deeper knowledge of the intrinsic community structure of networks, which has been largely ignored by previous methods. In this work, we integrate network topological information and node attribute information uniformly in a novel \({\ell }_{2,1}\)-norm nonnegative matrix factorization (NMF)-based model with network regularization. Specifically, we employ the \({\ell }_{2,1}\)-norm NMF to alleviate noise that exists in the node attributes, and incorporate a network regularization term into our framework to encode the network geometrical information. Accordingly, the new representations in the latent node space share the information encoded consistently in both the network structure and node attributes; giving the new representations greater discriminatory power. Our experimental results on six real-world networks demonstrate the competitive performance of the proposed method compared to the state-of-the-art approaches.
Authors: Zhuanlian Ding, Xihong Jiang, Xi Hu
Institution: Anhui University
Language: English
Subfield: Statistical and Nonlinear Physics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4412789986