B0380
Title: Bi-directional clustering via averaged mixture of finite mixtures
Authors: Tianyu Pan - University of California, Irvine (United States) [presenting]
Weining Shen - UC Irvine (United States)
Guanyu Hu - The University of Texas Health Science Center at Houston (United States)
Abstract: Bi-directional clustering is an approach that captures the heterogeneity of a data matrix on both rows and columns simultaneously. It has been widely applied in a variety of fields, such as genomics, economics, sports, etc., to detect the clustering effect on variable-level (column) and subject-level (row) accordingly. Yet it remains under-discovered whether such bi-directional heterogeneity can be well defined and proved to be effective using a statistical model. A density-based bi-directional clustering approach is proposed by averaging over a mixture of finite mixture models (MFMs), termed an averaged mixture of finite mixtures. The model has the proven ability to capture such heterogeneity asymptotically and provide root n rate (up to a log term) contracted estimations on both density and parameters. The proposed model is manifested to be effective and tractable using simulations and helpful in mining the statistical dependency between random variables based on the applications to a Georgia county economic dataset.
