Semi-supervised Discriminant Analysis with Tensor Representation

Discriminant Analysis with Tensor Representation (DATER) is a typical supervised tensor embedding methods which have been used in face recognition and behavior recognition. Commonly, the projection matrixes are obtained by maximizing the interclass scatters and simultaneously minimizing the intraclass scatters. But in practical applications, no sufficient labeled training samples with prior knowledge was provided, so unlabeled image data are eager for incorporating in subspace learning algorithm to improve the identification accuracy. In this paper, we propose a tensor space learning method, which are called Semi-supervised Discriminant Analysis with Tensor Representation (SDATER). The algorithm maximizes the separability of different classes using DATER algorithm, while preserving the local topological manifold structure and the intrinsic geometric property using Tensor Neighborhood Preserving Embedding (TNPE). Therefore, our SDATER algorithm can introduce high order tensor discriminant analysis into simi-supervised learning and avoid the curse of dimensionality problem and the small sample size problem. Experimental results demonstrate that our SDATER algorithm outperform the start-of-art methods.