Olgica Milenkovic, University of Illinois, Urbana-Champaign, USA

Title:

New Approaches to Hypergraph Clustering and Community Detection

Abstract:

Motivated by emerging problems in network analysis, learning hierarchical rankings and image segmentation, we introduce a new paradigm in hypergraph clustering and community detection termed hypergraph correlation clustering. Hypergraph correlation clustering is an agnostic combinatorial learning method which allows for capturing communities with a priori unknown structures based only on the knowledge of hypergraph edge cardinalities. The method has statistical counterparts which can be applied to hypergraph generalizations of stochastic block models and for other community detection paradigms.