Summarization of Static Graphs

Abstract

One natural way to understand a graph and its underlying processes is to visualize and interact with it. However, for large datasets with several millions or billions of nodes and edges, such as the Facebook social network, even loading them using an appropriate visualization software requires significant amount of time. If the memory requirements are met, visualizing the graph is possible, but the result is a “hairball” without obvious patterns: often the number of nodes is larger than the number of pixels on a screen, while, at the same time, people have limited capacity for processing information. How can we summarize efficiently, and in simple terms, which parts of the graph stand out? What can we say about its structure? Its edge distribution will likely follow a power law [64], but apart from that, is it random? The focus of this chapter is finding short summaries for large graphs, in order to gain a better understanding of their characteristics.