Discriminative cost sensitive Laplacian score for face recognition

In recent years, face recognition is being recognized as a cost sensitive learning problem. For example, in a door-locker based on the face recognition system, it may make a gallery person inconvenient, who is misrecognized as an impostor and not allowed to enter the room, but it could result in a serious loss or damage if an imposter is misrecognized as a gallery person and allowed to enter the room. To deal with the cost sensitive problem in face recognition, many cost sensitive classifiers have been proposed. However, face recognition is a high dimensional problem, no sufficient attention is paid to the research on cost sensitive feature selection. In this paper, we propose a cost sensitive feature selection method called Discriminative Cost Sensitive Laplacian Score (DCSLS) for face recognition. The main contributions of DCSLS are as follows: (1) DCSLS incorporates the idea of local discriminant analysis into Laplacian Score, which prefers the features that can minimize the local neighborhood relationship of within-class and maximize the local neighborhood relationship of between-class, simultaneously; (2) DCSLS embeds the misclassification cost in Laplacian Score, which satisfies the minimal misclassification loss criterion. Extensive experimental results on six face data sets: ORL, Extended Yale B, PIE, AR, FERET and FRGC-204 show the superiority of DCSLS.