An Improved Cost-Sensitive Learning Algorithm for Face Recognition

Currently cost-sensitive learning has become one of hotspots and has been applied to the fields of pattern recognition to solve related problems recently. However, the algorithm which applies it to the stage of feature extraction is relatively rare. Though Jiwen Lu´s work is relatively novel, which is still applied to the stage of classification in pursuit of the minimum cost of the overall classification error. Therefore it is not for the purpose of high recognition rate which pattern recognition requires. In this paper cost-sensitive learning is applied to the stage of feature extraction of the face recognition successfully, and we present a novel improved cost-sensitive learning method. We report experiments on the AR database which demonstrates that the proposed method dramatically improves the recognition rate relative to linear discriminant analysis and locality preserving projections.