局部系数增强判别协同表示分类法  

Local coefficient enhanced discriminant collaborative representation classification method

作　　者：简彩仁[1] 夏靖波[1] JIAN Cairen;XIA Jingbo(Tan Kah Kee College,Xiamen University,Zhangzhou 363105,China)

机构地区：[1]厦门大学嘉庚学院,福建漳州363105

出　　处：《厦门大学学报（自然科学版）》2023年第3期487-492,共6页Journal of Xiamen University：Natural Science

基　　金：福建省自然科学基金(2020J01039);福建省中青年教师教育科研项目(科技类)(JAT210631)。

摘　　要：协同表示分类法(CRC)是一种基于表示理论的简单分类方法,它和它的许多扩展模型具有良好的分类性能.概率协同表示分类法(ProCRC)、判别稀疏表示分类法(DSRC)和判别协同表示分类法(DCRC)通过定义不同的判别信息项改进CRC.借鉴ProCRC、DSRC和DCRC,定义判别信息项增强表示系数的判别能力,基于近邻系数的相似性,定义局部系数强化项,增强近邻样本的协同表示能力,提出了局部系数增强判别协同表示分类法(LDCRC).在6个数据集上的实验表明LDCRC是有效的.As a simple classification method based on the representation theory,collaborative representation classification(CRC)along with a great number of its extended models have performed satisfactorily.Also,probabilistic collaborative representation classification(ProCRC),discriminant sparse representation classification(DSRC)and discriminant collaborative representation classification(DCRC)define different discriminant information items to improve CRC.Using experiences of these three RCs for references,we define the discriminant information item to enhance the discriminant ability of representation coefficient.Based on the similarity of neighborhood coefficients,the local coefficient enhancement term is defined to enhance the collaborative representation ability of neighborhood samples.Therefore,the local coefficient enhanced discriminant collaborative representation classification method(LDCRC)is proposed.Experiments on six data sets show the effectiveness of LDCRC.

关 键 词：协同表示 局部系数 判别 分类 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP371[自动化与计算机技术—计算机科学与技术]