A Robust Approach for Multivariate Binary Vectors Clustering and Feature Selection

摘要

Given a set of binary vectors drawn from a finite multiple Bernoulli mixture model, an important problem is to determine which vectors are outliers and which features are relevant. The goal of this paper is to propose a model for binary vectors clustering that accommodates outliers and allows simultaneously the incorporation of a feature selection methodology into the clustering process. We derive an EM algorithm to fit the proposed model. Through simulation studies and a set of experiments involving handwritten digit recognition and visual scenes categorization, we demonstrate the usefulness and effectiveness of our method.