Entropy-Based Variational Learning of Finite Generalized Inverted Dirichlet Mixture Model

Mohammad Sadegh Ahmadzadeh, Narges Manouchehri, Hafsa Ennajari, Nizar Bouguila, Wentao Fan

Abstract: Mixture models are considered as a powerful approach for modeling complex data in an unsupervised manner. In this paper, we introduce a finite generalized inverted Dirichlet mixture model for semi-bounded data clustering, where we also developed a variational entropy-based method in order to flexibly estimate the parameters and select the number of components. Experiments on real-world applications including breast cancer detection and image categorization demonstrate the superior performance of our proposed model.