On the schatten norm for matrix based subspace learning and classification

Schatten norm, especially nuclear norm (p=1) has been widely used as an approximation of matrix rank and regularized term in the criterion function in pattern recognition and machine learning. In this paper, we point out that Schatten norm (p≤1) is also an effective and robust distance metric in the classification stage and can help improve the classification accuracy of matrix based feature extraction methods. Extensive experiments illustrate the effectiveness of Schatten norm (p≤1).