Non-negative Sparse and Collaborative Representation for Pattern Classification
Sparse representation (SR) and collaborative representation (CR) have been successfully applied in many pattern classification tasks such as face recognition. In this paper, we propose a novel Non-negative Sparse and Collaborative Representation (NSCR) for pattern classification. The NSCR representation of each test sample is obtained by seeking a non-negative sparse and collaborative representation vector that represents the test sample as a linear combination of training samples. We observe that the non-negativity can make the SR and CR more discriminative and effective for pattern classification. Based on the proposed NSCR, we propose a NSCR based classifier for pattern classification. Extensive experiments on benchmark datasets demonstrate that the proposed NSCR based classifier outperforms the previous SR or CR based approach, as well as state-of-the-art deep approaches, on diverse challenging pattern classification tasks.