Motif-based embedding for graph clustering. (1st December 2016)

• Record Type: Journal Article
• Authors: Lim, Sungsu; Lee, Jae-Gil
• Is Part Of: Journal of statistical mechanics. (2016:Dec.)
• Publication Date: 2016-12-01
• Subjects: 11; Statistical mechanics -- Periodicals; Mechanics -- Statistical methods -- Periodicals; 530.1305
• DOI: 10.1088/1742-5468/2016/12/123401
• Languages: English
• ISSNs: 1742-5468

Abstract

Abstract: Community detection in complex networks is a fundamental problem that has been extensively studied owing to its wide range of applications. However, because community detection methods typically rely on the relations between vertices in networks, they may fail to discover higher-order graph substructures, called the network motifs . In this paper, we propose a novel embedding method for graph clustering that considers higher-order relationships involving multiple vertices. We show that our embedding method, which we call motif-based embedding, is more effective in detecting communities than existing graph embedding methods, spectral embedding and force-directed embedding, both theoretically and experimentally.