基于子空间学习和伪标签回归的无监督特征选择  

Subspace Learning and Virtual Label Regression Based Unsupervised Feature Selection

作　　者：盛超 宋鹏 郑文明[2] 赵力[3] SHENG Chao;SONG Peng;ZHENG Wenming;ZHAO Li(School of Computer and Control Engineering,Yantai University,Yantai,Shandong 264005,China;Key Laboratory of Child Development and Learning Science of Ministry of Education,Southeast University,Nanjing,Jiangsu 210096,China;School of Information Science and Engineering,Southeast University,Nanjing,Jiangsu 210096,China)

机构地区：[1]烟台大学计算机与控制工程学院,山东烟台264005 [2]东南大学儿童发展与学习科学教育部重点实验室,江苏南京210096 [3]东南大学信息科学与工程学院,江苏南京210096

出　　处：《信号处理》2021年第9期1701-1708,共8页Journal of Signal Processing

基　　金：国家自然科学基金(61703360);中央高校基本科研业务费专项资金(2242021k30014,2242021k30059)。

摘　　要：信息技术的快速发展产生了大量无标签高维数据。为了能够更好地处理这些数据,提出了一种基于子空间学习和伪标签回归的无监督特征选择方法。首先,从矩阵分解的角度将子空间学习和特征选择结合在一个框架中,并且对特征选择矩阵施加L 2,1范数保证稀疏,在寻找原始数据空间低维表示的同时进行特征选择;其次,利用回归函数来学习特征子空间和伪标签之间的映射关系,利用伪标签和回归函数来指导无监督特征选择,以使选择出来的特征更具判别力;最后,通过引入图拉普拉斯来挖掘隐藏在样本空间和特征空间的局部结构信息。在六个公开的数据集上进行了实验,实验结果表明该方法要优于其他几种先进的无监督特征选择算法。With the rapid development of information technology,a lot of unlabeled high-dimensional data are generated.To cope with these data,in this paper,we proposed a subspace learning and virtual label regression based unsupervised feature selection method.First,from the viewpoint of matrix factorization,we combined subspace learning and feature selection into a joint framework and constrained the feature selection matrix with an L 2,1-norm to select features while finding the low-dimensional representation of the original data space.Then,we utilized the regression function to learn the mapping relationship between the feature subspace and the virtual label space.With the guide of virtual label matrix and regression function,we can select the discriminative features.Finally,we introduced a graph Laplacian to explore the local information hidden in the data space and feature space.We conducted extensive experiments on six public datasets.The results show that our method is superior to some state-of-the-art unsupervised feature selection methods.

关 键 词：子空间学习 伪标签回归 图正则 局部结构保持 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]