基于结构和约束保持的半监督特征选择  

Semi-supervised feature selection based on structure and constraints preserving

作　　者：潘俊[1] 王瑞琴[2] 孔繁胜[3] 

机构地区：[1]温州大学信息安全研究所,浙江温州325035 [2]温州大学物理与电子信息工程学院,浙江温州325035 [3]浙江大学计算机科学与技术学院,浙江杭州310027

出　　处：《南京理工大学学报》2014年第4期518-525,共8页Journal of Nanjing University of Science and Technology

基　　金：浙江省科技计划项目(2012C33086);浙江省自然科学基金(LQ12F02008)

摘　　要：针对现有特征选择算法较少同时考虑样本的空间结构和先验知识的不足,提出一种基于结构和约束保持的半监督特征选择方法。该方法采用成对约束作为先验知识,同时考虑局部和非局结构,定义了一种新的特征评价准则——结构和约束保持分值。利用大量的无标记样本来学习样本空间结构,利用少量的成对约束信息来学习类内和类间边缘,所选择的特征子集能较好地保持空间结构信息和类属信息。在多个数据集上的实验结果表明,和现有的几种特征排序选择算法相比,所提方法有较好表现。To overcome the deficiency of most existing feature selection methods which fairly respect both the geometrical structure and the supervision information, a novel approach called semi-supervised feature selection based on structure and constraints preserving is proposed. In this method,both the pairwise constraints and the local and nonlocal structure are taken into account,and a new feature selection criterion,i. e. structure and constraints preserving（ SCP） score is defined. The SCP score exploites abundant unlabeled data points to learn the geometrical structure of the data space,and uses a few pairwise constraints to discover the margins of different classes. Those features that can preserve the geometrical structure and pairwise constraints information are selected. Experimental results from several datasets show that the proposed method achieves better performance than the feature ranking selection methods.

关 键 词：特征选择 半监督学习 成对约束 结构和约束保持 特征排序 空间结构 先验知识 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]