Rene Gutierrez, Aaron Scheffler, Rajarshi Guhaniyogi, Abigail Dickinson , Charlotte DiStefano, and Shefali Jeste
08/24/2022 08:57 AM
Statistics
Clustering of tensors with limited sample size has become prevalent in a variety of application areas. Existing Bayesian model based clustering of tensors yields less accurate clusters when the tensor dimensions are sufficiently large, sample size is low and clusters of tensors mainly reveal difference in their variability. This article develops a clustering technique for high dimensional tensors with limited sample size when the clusters show difference in their covariances, rather than in their means. The proposed approach constructs several matrices from a tensor, referred to as transformed features, to adequately estimate its variability along different modes and implements a model-based approximate Bayesian clustering algorithm with the matrices thus constructed, in place with the original tensor data. Although some information in the data is discarded, we gain substantial computational efficiency and accuracy in clustering. Simulation study assesses the proposed approach along with its competitors in terms of estimating the number of clusters, identification of the modal cluster membership and the probability of mis-classification in clustering (a measure of uncertainty in clustering). The proposed methodology provides novel insights into potential clinical subgroups for children with autism spectrum disorder based on resting-state electroencephalography activity.
