Seminar: An Analytic Approach to Community Detection in Directed Weighted Graphs

Speaker: Priya Bucha Jain, Western University

Title: An Analytic Approach to Community Detection in Directed Weighted Graphs

Abstract: Network data has applications across biology, physics, ecology, and many other fields. A common feature of networks across these fields is community structure, i.e., the arrangement of nodes into clusters with strong connections. Detecting community structure in network data is an important problem, and many computational algorithms have been developed for this application. In this talk, I will present a new analytical approach for detecting community structure in directed, weighted networks. This approach relies solely on the eigen spectrum of the network adjacency matrix and does not entail free parameters to be tuned or trained. I will describe this approach and provide a method for estimating the number of communities. I will prove the validity of this approach in an ideal case and then demonstrate its efficacy through numerical simulations in weighted graphs with randomness on individual edges.