Abstract
Complex network is a mainstream form of unstructured data in real world. Detecting communities in complex networks bears a wide range of applications. Different from the existing methods, which concentrate on applying statistics, graph theory or combinations, this work presents a new algorithm along a geometric avenue. By utilizing normalized discrete Ricci flow with modified sigma-weight-sum, and employing a limit-free Ricci curvature using *-coupling, this algorithm prevents the graph from collapsing to a point, and eliminates a hyper parameter alpha in discrete Ollivier Ricci curvature. Besides, experiments on real-world networks and artificial networks have shown that this normalized algorithm has a matching or better result, and is more robust with regard to unnormalized one (Ni et al., 2019). The code is available at https://github.com/laiguzi/NormalizedRicciFlow. (C) 2022 Elsevier B.V. All rights reserved.
NSTL
Elsevier