Graph Compression

Abstract

Graphs form the foundation of many real-world datasets. At the same time, the size of graphs presents a big obstacle to understand the essential information they contain. In this report, I mainly review the framework in article [1] for compressing large graphs. It can be used to improve visualization, to understand the high-level structure of the graph, or as a pre-processing step for other data mining algorithms. The compression model consists of a graph summary and a set of edge corrections. This framework can produce either lossless or lossy compressed graph representations. Combined with Minimum Description Length (MDL), it can produce a compact summary. The performance of this framework is evaluated on multiple sets of real data graph.