Supervised Probabilistic Principal Component Analysis Mixture Model in Dimensionality Reduction without Loss Framework for Face Recognition

In this study, first a supervised version for probabilistic principal component analysis mixture model (SPPCAMM) is proposed. Then, considering projection penalty in learning of a predictive model, a method for face recognition using a dimensionality reduction without loss framework is proposed. In the proposed method, first a locally linear underlying manifold of data samples is obtained using supervised probabilistic principal component analysis mixture model. Then, a support vector machine with projection penalty is trained as the mentioned predictive model using this locally linear underlying manifold. In this way, the benefits of dimensionality reduction are used in the predictive model, while using the projection penalty idea, the loss of useful information is prevented. To train and evaluate the proposed method, well-known face databases are used. Gabor feature extraction method is applied on the face images. The experimental results show that the proposed method has the most average classification accuracy compared to many traditional methods, and also compared to the projection penalty idea used for linear and non-linear kernel-based dimensionality reduction techniques.