Fast SimRank Computation over Disk-Resident Graphs

Abstract

There are many real-world applications based on similarity between objects, such as clustering, similarity query processing, information retrieval and recommendation systems. SimRank is a promising measure of similarity based on random surfers model. However, the computational complexity of SimRank is high and several optimization techniques have been proposed. In the paper optimization issue of SimRank computation in disk-resident graphs is our primary focus. First we suggest a new approach to compute SimRank.Then we propose optimization techniques that improve the time cost of the new approach from O (kN ² D ²) to O(kNL), where k is the number of iteration, N is the number of nodes, L is the number of edges, and D is the average degree of nodes. Meanwhile, a threshold sieving method is presented to reduce storage and computational cost. On this basis, an external algorithm computing SimRank in disk-resident graphs is introduced. In the experiments, our algorithm outperforms its opponent whose computation complexity also is O(kNL).

This work is supported by the Fundamental Research Funds for the Central Universities,and the Research Funds of Renmin University of China(Grant No.12XNH178).