Exploring the Distribution of Local Neighborhood Structures in Large Networks

Network exploration techniques aim at identifying features like clusters, central nodes or motifs in networks. Motifs are local sub-networks of a network that appear more frequently than expected. We introduce a new type of local neighborhood structure that is spectrally defined, i.e., based on the eigenvalues and eigenvectors of some matrix associated with the network, and show that well established visual analytics techniques for exploring high-dimensional point clouds provide an effective means for the exploration of the distribution of these neighborhood structures. Experiments on real world social networks demonstrate that our approach is indeed capable of revealing interesting neighborhood structures that are not easily accessible otherwise.