作者: Jaekoo Lee , Gunn Kim , Sungroh Yoon
关键词:
摘要: By how much is a large-scale graph transformed over time or by significant event?' 'how structurally similar are two graphs?' the questions that this paper attempts to address. The proposed method efficiently calculates and accurately produces similarity. Our approach based on well-known random walk with restart (RWR) algorithm, which quantifies relevance between nodes express structural connection characteristics of graphs. Intergraph compression, inspired interframe merges input graphs reorders their contributing improved process-data storage efficiency processing convenience. This boon RWR algorithm for representation via intergraph compression can be used show similarity because sub-matrix blocks reordered concentrate nonzero elements. In performing inter-node relevance, efficient in space requirement results more quickly conventional transformation schemes. We demonstrate validity our through experiments apply it usage data public transportation SmartCard large metropolitan area suggest usefulness algorithm.