A simple approach to bi-clustering discrete data
Marco Alfò, Francesca Martella, Mario Marino · IRIS Research product catalog (Sapienza University of Rome) · 2015
Finite Mixtures of Factor Analyzers have been used as model based clustering approach for high-dimensional data. In the last few years, they have been extended to biclustering purposes or to deal with Gaussian and non-Gaussian, continuous, responses, for example by using t and generalized hyperbolic distributions. We propose an extension of Finite Mixtures of Factor Analyzers to allow for simultaneous clustering of subjects and variables when multivariate discrete outcomes are available. We detail the EM algorithm for ML parameter estimation and discuss the performance of the proposed model when applied to binary synthetic and to real count data, coming from Next-generation sequencing assays.