Techniques for clustering interaction data as a collection of graphs

摘要: A natural approach to analyze interaction data of form "what-connects-to-what-when" is create a time-series (or rather sequence) graphs through temporal discretization (bandwidth selection) and spatial (vertex contraction). Such together with non-negative factorization techniques can be useful for obtaining clustering graphs. Motivating application performing (as opposed vertex clustering) found in neuroscience social network analysis, it also used enhance community detection (i.e., by way conditioning on the cluster labels. In this paper, we formulate problem as model selection problem. Our involves information criteria, matrix singular value thresholding, illustrate our using real simulated data.