Variable Selection for Clustering Three-Way Data

Ample work on dimension reduction for multivariate model-based clustering has been conducted; however, to date, relatively few dimension reduction methods have been presented in the matrix-variate paradigm. Such work is, for example, useful for modelling data arising from longitudinal studies with multiple responses, multivariate repeated measures data, or image data. Similar to the multivariate paradigm, issues persist when clustering data with noisy and uninformative variables within the matrix-variate paradigm. Thus, a variable selection algorithm for the matrix-variate paradigm is presented and tested on real datasets.