A0631
Title: Heterogeneous block covariance model for community detection
Authors: Xiang Li - The George Washington University (United States) [presenting]
Yunpeng Zhao - Colorado State University (United States)
Qing Pan - George Washington University (United States)
Ning Hao - University of Arizona (United States)
Abstract: Community detection is a clustering method based on objects' pairwise relationships such that objects classified in the same group are more densely connected than objects from different groups. Most of the model-based community detection methods such as the stochastic block model and its variants are designed for networks in which the connections between nodes are described by discrete values, which ignores the practical scenarios where the pairwise relationships between nodes can be continuous. The heterogeneous block covariance model (HBCM) proposes a novel clustering structure applicable to signed and continuous connections between nodes such as a covariance matrix, taking into account the characteristics of each individual object additional to the group-level information, and uses the variational EM algorithm to estimate the optimal group membership and parameters. The statistical property of the HBCM is studied and its practical performance is demonstrated by extensive numerical simulations. The HBCM is applied to the yeast gene expression data.