A Higher-Order Latent Space Network Model

Nesreen K. Ahmed, Ryan A. Rossi, Theodore L. Willke, Rong Zhou

Abstract: Previous work in network analysis has focused on modeling node roles in the graph. In this work, we introduce edge role discovery and develop a general framework for modeling edge roles in large networks. In addition, a general class of higher-order role discovery methods are proposed that leverage features based on induced subgraphs (graphlets, motifs) for learning better and more useful roles. All methods are fast with a runtime that is linear in the number of edges and able to scale to large real-world networks via an effective parallel implementation. The experimental results demonstrate the utility of edge roles for network analysis tasks on a variety of graphs from various problem domains.