A0634
Title: Community detection using variational EM approaches
Authors: Yunpeng Zhao - Colorado State University (United States) [presenting]
Qing Pan - George Washington University (United States)
Ning Hao - University of Arizona (United States)
Xiang Li - The George Washington University (United States)
Abstract: Community detection involves clustering objects based on their pairwise relationships. Two problems are addressed in community detection: Clustering in bipartite graphs and the heterogeneous block covariance model (HBCM) for weighted graphs. In both scenarios, the standard expectation-maximization algorithm is computationally intractable, as the e-step requires summing over an exponentially large number of terms. To address this, an efficient variational EM algorithm is developed to estimate group membership. The methods are applied to a single-cell RNA-seq dataset from a mouse embryo and a movie ratings dataset.
