Maximum Margin Learning Projections for Face Recognition

This paper presents a novel dimensionality reduction algorithm called maximum margin learning projections (MMLP) for face recognition. MMLP exploits the geometrical and discriminant structures of the data points. In this way, MMLP can seek the subspace which optimally preserves the local neighborhood information of the data set and maximizes the margin between data points from different classes in each local area. Experimental results on the ORL and Yale face databases demonstrate that MMLP outperforms most of the state-of-the-art methods.