CONVERDENSE: Framework of Convergence Summary Graph for Large Numbers of Networks

Abstract

The recent development of high-throughput technologies provides a range of opportunities to systematically characterize diverse types of large networks. Large Network has been an emerging field in graph mining. This is happen since all of interaction between components formally simulated using graph. A potential solution to the false pattern problem is mining frequent subgraphs directly. A subgraph is frequent if it occurs multiple times in a set of
graphs. However, these approaches encounter scalability and interpretability issues when being applied to massive networks: (1) The number of frequent dense subgraphs is explosive when there are very large frequent dense subgraphs, e.g. subgraphs with hundreds of edges. (2) A frequent dense subgraph may not represent a tight association among its nodes. We develop a novel algorithm, to mine convergence dense subgraphs, a concept having better interpretability than frequent graph. All edges in a convergence subgraph should exhibit correlated occurrences in the whole graph set.