基于自适应Lasso流形规整的特征提取算法研究  

Research of feature extraction algorithm based on adaptive Lasso manifold regularization

作　　者：袁宝红[1] 卢宇 胡婷芳[2] Yuan Baohong;Lu Yu;Hu Tingfang(School of Electronic and Electrical Engineering,Anhui Sanlian University,Hefei 230601,China;The Office of Academic Studies,Anhui University,Hefei 230601,China)

机构地区：[1]安徽三联学院电子电气工程学院,安徽合肥230601 [2]安徽大学教务处,安徽合肥230601

出　　处：《湖南文理学院学报（自然科学版）》2021年第4期23-26,共4页Journal of Hunan University of Arts and Science(Science and Technology)

基　　金：安徽省教育厅高校自然科学项目(KJ2018A0600);安徽三联学院平台重点研究项目(PTZD2021019);安徽三联学院校级科研项目(KJYB2019002,PTZD2020006)。

摘　　要：针对原始高维空间数据特征冗余的特征问题,提出了一种自适应Lasso流形规整的特征提取方法。在原始空间中的样本,经过投影后可以保持在原始空间中的近邻结构,投影到低维空间后也可像高维空间中那样相近。通过模型做完特征选择后,以这些被选的特征子集作为输入,在数据集中做了一系列的分类实验。结果表明,该算法可以精准提取高维样本集的低维流形结构,具有较小的尺寸误差递减,特征提取性能较好。Aiming at the original high-dimensional space problem,an adaptive Lasso manifold regularization feature extraction method is proposed.The sample in the original space can maintain the neighbor structure in the original space after being projected,and can also be as close as in the high-dimensional space after being projected into the low-dimensional space.After the feature selection is done through the model,a series of classification experiments are done in the data set with these selected feature subsets as input.Experiments show that the algorithm can accurately extract the low-dimensional manifold structure of the high-dimensional sample set,with small size error reduction,and good feature extraction performance.

关 键 词：自适应Lasso 流形规整 特征提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]