Regularized Complete Linear Discriminant Analysis for Small Sample Size Problems
In small sample size (SSS) problems, the number of available training samples is smaller than the dimensionality of the sample space. Since linear discriminant analysis (LDA) requires the within-class scatter matrix to be non-sigular, LDA cannot be directly applied to SSS problems. In this paper, regularized complete linear discriminant analysis (RCLDA) is proposed to solve SSS problems. RCLDA uses two regularized criterion to derive “regular” discriminant vectors in the range space of the within-class scatter matrix and “irregular” discriminant vectors in the null space of the within-class scatter matrix. Extensive experiments on the SSS problem of face recognition are carried out to evaluate the proposed algorithm in terms of classification accuracy and demonstrate the effectiveness of the proposed algorithm.
KeywordsComplete linear discriminant analysis regularization small sample size problems
Unable to display preview. Download preview PDF.