一种新的L1度量Fisher线性判别分析研究-计算机工程与应用.PDF

128 2018 ，54（4 ） Computer Engineering and Applications 计算机工程与应用 ⦾模式识别与工人智能⦾ 用 一种新的 度量Fisher 线性判别分析研究 L1 应 余景丽，胡恩良，张 涛 与 YU Jingli, HU Enliang, ZHANG Tao程 g 云南师范大学 数学学院，昆明 650500 r 工 o School of Mathematics, Yunnan Normal University, Kunming 650500, China . 机 j a L YU Jingli, HU Enliang, ZHANG Tao. Study of Fisher linear discriminant analysis based on 1 -norm. Computer 算 e Engineering and Applications, 2018, 54 （4 ）：128-134. c 计 . Abstract ：Fisher Linear Discriminant Analysis （FLDA ）is a classical method of feature extraction with supervised infor- w mation, which maximizes the Fisher criterion to find the optimal projection matrix. In the criterion of standard FLDA, the w involved metric is based on L2 norm metric, which is usually lack of robustness and sensitive to outliers. In order to w improve the robustness, this paper proposes a new model and algorithm for FLDA, which is based on L1 norm metric. The experimental results show that, FLDA with L1 norm outperforms that with L2 norm in classification accuracy and robustness in many cases. Key words ：Fisher linear discriminant analysis; Fisher criterion; L1 norm metric; robustness; feature extraction 摘 要：Fisher 线性判别分析（Fisher Linear Discriminant Analysis ，FLDA ）是一种典型的监督型特征提取方法，旨在 L 最大化Fisher 准则，寻求最优投影矩阵。在标准Fisher 准则中，涉及到的度量为 范数度量，