A family of mixture models for biclustering

Biclustering is used for simultaneous clustering of the observations and variables when there no group structure known a priori. It being increasingly in bioinformatics, text analytics, so on. Previously, biclustering has been introduced model-based framework by utilizing similar to mixture factor analyzers. In such models, observed are modeled using latent variable that assumed be from . Clustering imposing constraints on entries loading matrix 0 1 results block diagonal covariance matrices. However, this approach overly restrictive as off-diagonal elements blocks matrices can only which lead unsatisfactory model fit complex data. Here, where matrix. This ensures terms within non-zero not restricted 1. leads superior A family models developed components For parameter estimation, an alternating expectation conditional maximization (AECM) algorithm used. Finally, proposed method illustrated simulated real datasets.