A0385
Title: Community detection in multilayer degree-corrected stochastic blockmodels
Authors: Joshua Agterberg - Johns Hopkins University (United States) [presenting]
Jesus Arroyo - Texas A&M University (United States)
Zachary Lubberts - University of Virginia (United States)
Abstract: In multilayer network analysis, networks often share some underlying structure (such as communities) but have possibly heterogeneous network-specific idiosyncrasies that can make simple averaging procedures fail even at the population level. We propose the multilayer degree-corrected stochastic blockmodel, a multilayer network model that assumes that communities are shared across networks, but that edge probabilities and degree corrections can vary between networks. First, we discuss why averaging procedures may fail, and then we discuss the underlying shared spectral geometry. Inspired by this geometry, we propose an algorithm to take advantage of the shared structure amongst the networks, and we show its performance on real and simulated data.
