Community detection based on information dynamics

Abstract

Community detection, as the primary task of network analysis, provides a promising way to summarize the network structure, study the interactions of groups and obtain insight into the potential network functions. Many community detection algorithms have been proposed from different viewpoints. However, uncovering the intrinsic communities in networks is still a non-trivial task due to the difficulty of parameter tuning, user-bias criteria and the lack of ground-truth information. In this paper, we propose a new dynamic model that aims to uncover the communities in networks in a more intuitive and topologically driven way. The basic idea is to explore the community structure by simulating the information exchange in a given network. The information that is flowing on a given network is governed by the underlying network community structure. The community structure, in turn, is reflected by the information dynamics. To prove the convergence of the information dynamic model, the monotone convergence theorem is employed. In the experiments, we have demonstrated that our proposed approach is superior to many representative algorithms on both synthetic and real-world networks.