S2R&R2S: A framework for ranking vertex and computing vertex-pair similarity simultaneously

Abstract

With the popularity of information networks and their wide range of applications, two fundamental operations in information network analysis, ranking vertices and computing similarity between vertices, are attracting growing interest among researchers. However, the existing studies about the two operations have been carried out independently and the relation between them has been ignored. In fact, there is a definite relationship between the vertex rank and vertex-pair similarity. In this paper, we propose a novel framework, S2R&R2S, to rank vertices and compute vertex-pair similarity simultaneously. The framework is composed of two methods: S2R and R2S. The S2R method ranks vertices with the help of vertex-pair similarity, and the R2S method computes vertex-pair similarity combining the network structure and the rank scores of vertices. In the proposed framework, S2R and R2S are performed iteratively. The experimental results demonstrate the effectiveness of the proposed approach in both vertex rank and similarity computation.