Non-greedy Max-min Large Margin based on L1-norm

摘要

In recent years, there have been several LI-norm-based linear projection methods and max-min-based dimensionality reduction methods, which show robustness to outliers and noises and show large margin for discrimination. In this paper, we propose LI-norm-based max-min large margin (MLM-L1) for linear projection-based dimensionality reduction. It makes use of the robustness of L1-norm to outliers and noises and the max-min idea for large margin. A non-greedy iterative algorithm (NMLM-L1) is proposed to solve the optimization problem of the proposed MLM-L1. Experiments on several face image databases show that the proposed method has better classification performance than its closely related methods. (C) 2018 Elsevier B.V. All rights reserved.