An unsupervised kernel optimization in dimensional reduction

Subspace analysis is an effective dimensional reduction approach for face recognition. Finding a suitable low dimensional subspace is a key step of subspace analysis, for it has a direct effect on recognition performance. In this paper, we propose a new subspace analysis method called center kernel unsupervised discriminant projection (CKUDP). The kernel trick is adopted to allow the efficient computation of unsupervised discriminant projection in high-dimensional feature space. Moreover, a center solution for obtaining the optimal feature vectors in feature space is presented which can preserve the most discriminative information. Experiments results on the ORL database and Yale database demonstrate the utility of the proposed approach.