Deep Discriminant Non-negative Matrix Factorization Method for Image Clustering

Non-negative matrix factorization (NMF) is a single-layer decomposition algorithm for image data analysis and cannot reveal the intrinsic hierarchical-structure information hidden under the data. Moreover, it extracts features with weak discriminant power due to merely using unlabeled data. These flaws cause the undesirable performance of NMF in image clustering. To address the problems of NMF, this paper presents a novel deep discriminant NMF (DDNMF) approach via neural networks (NN). Our DDNMF model is established by employing the loss function of neural networks and the supervised regularization terms. The optimization problem of DDNMF is solved utilizing gradient descent method (GDM). The proposed DDNMF algorithm is proven to be convergent and stable. Some related NMF-based algorithms are selected for comparison on two facial image datasets. The experimental results confirm the effectiveness and superior performance of our DDNMF method in image clustering tasks.