Description:

In studies of networks, researchers often examine the evolution of mesoscale structures, structures that involve groups of nodes that are larger than a single node but smaller than an overall network. In this talk, we present our use of statistical-inference methods to detect two such mesoscale structures, community structure and core—periphery structure, in time-dependent networks (i.e., “temporal networks”). We devise statistical-inference methods that avoid common biases in such methods against generating communities or other groups with large or small numbers of nodes. We show that our methods are able to accurately identify mesoscale structure in cases of interest. Additionally, we show that using our generative model is beneficial for analyzing the community structure of networks with large or small communities. It leads to better accuracy than methods that contain biases against generating groups with large or small numbers of nodes.

Theodore Faust, Postdoctoral Scholar, Southern Methodist University

Host: Istvan Kovacs