B1079
Title: New directions in constrained spectral clustering for networks
Authors: Sandipan Roy - University of Bath (United Kingdom) [presenting]
Sinyoung Park - University of Bath (United Kingdom)
Matthew Nunes - University of Bath (United Kingdom)
Abstract: Spectral clustering has been used widely as a popular tool for community detection in data with network structure. Often there is additional information available for individuals in the networks that could be incorporated as constraints for community detection. Several possible techniques are explored to perform constrained spectral clustering in both sparse and dense networks. The theoretical properties of constrained spectral clustering are investigated under certain scenarios and the best possible way to incorporate them is highlighted. Additional covariate information and regularisation (sparse networks) methods were also considered in the constrained setting.