正在加载图片...
第11卷第3期 智能系统学报 Vol.11 No.3 2016年6月 CAAI Transactions on Intelligent Systems Jun.2016 D0I:10.11992/is.2016030. 网s络出版地址:http:/www.cnki.net/kcms/detail/23.1538.TP.20160513.0924.024.html 一种改进的投影孪生支持向量机 花小朋,孙一颗,丁世飞2 (1.盐城工学院信息工程学院,江苏盐城224051;2.中国矿业大学计算机学院,江苏徐州221116) 摘要:针对投影孪生支持向量机(PTSVM)在训练阶段欠考虑样本空间局部结构和局部信息的缺陷,提出一种具有 一定局部学习能力的有监督分类方法:加权投影孪生支持向量机(weighted PTSVM,WPTSVM)。相比于PTSVM, WPTSVM优势在于:通过构造类内近邻图为每个样本获取特定的权值,并且以加权均值取代标准均值,在一定程度 上提高了算法的局部学习能力:选取异类样本集中少量边界点构造优化问题的约束条件,很大程度上降低了二次规 划求解的时间复杂度:继承了PTSVM的优点,可以看成PTSVM的推广算法。理论分析及其在人造数据集和真实数 据集上的测试结果表明该方法具有上述优势。 关键词:分类:投影孪生支持向量机;局部信息;加权均值;近邻图;二次规划;约束条件;时间复杂度 中图分类号:TP391.4文献标志码:A文章编号:1673-4785(2016)03-0384-06 中文引用格式:花小朋,孙一颗,丁世飞.一种改进的投影孪生支持向量机[J].智能系统学报,2016,11(3):384-391. 英文引用格式:HUA Xiaopeng,SUN Yike,.DING Shifei..An improved projection twin support vector machine[J].CAAI transac- tions on intelligent systems,2016,11(3):384-391. An improved projection twin support vector machine HUA Xiaopeng',SUN Yike',DING Shifei2 (1.School of Information Engineering,Yancheng Institute of Technology,Yancheng 224051,China;2.School of Computer Science and Technology,China University of Mining and Technology,Xuzhou 221116,China) Abstract:A supervised classification method having a local learning ability,called weighted projection twin support vector machine(WPTSVM),is proposed.This method aims to improve upon a defect that projection twin support vector machines (PTSVMs)have,namely,that PTSVMs do not take account of the local structure and local infor- mation of a sample space in the training process.Compared with PTSVM,WPTSVM improves its local learning a- bility to some extent by attaching different weights for each sample according to the within-class neighborhood graph and replacing the standard mean with a weighted mean.Moreover,to reduce computational complexity,WPTSVM chooses a small number of boundary points in the contrary-class based on the between-class neighborhood graph to construct constraints of the original optimization problems.The method inherits the merits of PTSVM and can be re- garded as an improved version of PTSVM.Experimental results on artificial and real datasets indicate the effective- ness of the WPTSVM method. Keywords:classification;projection twin support vector machine;local information;weighted mean;neighborhood graph;quadratic programming;constraint condition;time complexity 在分类问题中,经典支持向量机(SVM)依据间 隔最大化准则生成分类决策面,存在训练时间复杂 度偏高和欠考虑样本分布情况的缺陷2。近年 收稿日期:2016-03-20.网络出版日期:2016-05-13. 来,作为经典SVM的改进方法,非平行超平面分类 基金项目:国家重点基础研究计划项目(2013CB329s02):国家自然科学器(nonparellel hyperplane classifiers,NHCs)[)已经 基金项目(61379101):江苏省自然科学基金项目 (BK20151299). 成为模式识别领域新的研究热点之一。孪生支持向 通信作者:花小朋.E-mail:xp_hua@163.com第 11 卷第 3 期 智 能 系 统 学 报 Vol.11 №.3 2016 年 6 月 CAAI Transactions on Intelligent Systems Jun. 2016 DOI:10.11992 / tis.2016030. 网络出版地址:http: / / www.cnki.net / kcms/ detail / 23.1538.TP.20160513.0924.024.html 一种改进的投影孪生支持向量机 花小朋1 ,孙一颗1 ,丁世飞2 (1.盐城工学院 信息工程学院,江苏 盐城 224051; 2. 中国矿业大学 计算机学院,江苏 徐州 221116) 摘 要:针对投影孪生支持向量机(PTSVM)在训练阶段欠考虑样本空间局部结构和局部信息的缺陷,提出一种具有 一定局部学习能力的有监督分类方法:加权投影孪生支持向量机(weighted PTSVM,WPTSVM)。 相比于 PTSVM, WPTSVM 优势在于:通过构造类内近邻图为每个样本获取特定的权值,并且以加权均值取代标准均值,在一定程度 上提高了算法的局部学习能力;选取异类样本集中少量边界点构造优化问题的约束条件,很大程度上降低了二次规 划求解的时间复杂度;继承了 PTSVM 的优点,可以看成 PTSVM 的推广算法。 理论分析及其在人造数据集和真实数 据集上的测试结果表明该方法具有上述优势。 关键词:分类;投影孪生支持向量机;局部信息;加权均值;近邻图;二次规划;约束条件;时间复杂度 中图分类号:TP391.4 文献标志码:A 文章编号:1673⁃4785(2016)03⁃0384⁃06 中文引用格式:花小朋,孙一颗,丁世飞.一种改进的投影孪生支持向量机[J]. 智能系统学报, 2016, 11(3): 384⁃391. 英文引用格式:HUA Xiaopeng, SUN Yike, DING Shifei. An improved projection twin support vector machine[J]. CAAI transac⁃ tions on intelligent systems, 2016, 11(3): 384⁃391. An improved projection twin support vector machine HUA Xiaopeng 1 , SUN Yike 1 , DING Shifei 2 (1. School of Information Engineering, Yancheng Institute of Technology, Yancheng 224051, China; 2. School of Computer Science and Technology, China University of Mining and Technology, Xuzhou 221116, China) Abstract:A supervised classification method having a local learning ability, called weighted projection twin support vector machine (WPTSVM), is proposed. This method aims to improve upon a defect that projection twin support vector machines (PTSVMs) have, namely, that PTSVMs do not take account of the local structure and local infor⁃ mation of a sample space in the training process. Compared with PTSVM, WPTSVM improves its local learning a⁃ bility to some extent by attaching different weights for each sample according to the within⁃class neighborhood graph and replacing the standard mean with a weighted mean. Moreover, to reduce computational complexity, WPTSVM chooses a small number of boundary points in the contrary⁃class based on the between-class neighborhood graph to construct constraints of the original optimization problems. The method inherits the merits of PTSVM and can be re⁃ garded as an improved version of PTSVM. Experimental results on artificial and real datasets indicate the effective⁃ ness of the WPTSVM method. Keywords:classification; projection twin support vector machine; local information; weighted mean; neighborhood graph; quadratic programming; constraint condition; time complexity 收稿日期:2016⁃03⁃20. 网络出版日期:2016⁃05⁃13. 基金项目:国家重点基础研究计划项目(2013CB329502);国家自然科学 基 金 项 目 ( 61379101 ); 江 苏 省 自 然 科 学 基 金 项 目 (BK20151299). 通信作者:花小朋. E⁃mail:xp_hua@ 163.com. 在分类问题中,经典支持向量机(SVM)依据间 隔最大化准则生成分类决策面,存在训练时间复杂 度偏高和欠考虑样本分布情况的缺陷[1⁃2] 。 近年 来,作为经典 SVM 的改进方法,非平行超平面分类 器( nonparellel hyperplane classifiers,NHCs) [3] 已经 成为模式识别领域新的研究热点之一。 孪生支持向
向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有