支持向量机的新发展.pdfVIP

第19 卷第5 期 　控　制　与　决　策 2004 年5 月 Vol. 19 No . 5 　Control and D ecis ion　 　 May 2004 : 1001-0920(2004) 05-0481-04 1 2 2 许建华 , 张学工, 李衍达 ( 1. , 210097; 2. , 100084) : Vapnik 等学者首先提出了实现统计学习理论中结构风险最小化原则 实用算法—支持向量机, 比较成功 地解决了模式分类问题. 其后, 机器学习界兴起了研究统计学习理论和支持向量机 热潮, 引人瞩目 研究分支有从 最优化技术出发改进或改造支持向量机, 依据统计学习理论和支持向量机 优点设计新 非线性机器学习算法等. 对此, 较为系统地回顾了近 10 年来算法研究领域 新发展. : 机器学习; 统计学习理论; 支持向量机 : T P18: A Advances in support vector machines 1 2 2 X U J ia n-hua , ZH A N G X ue-g ong , L I Ya n-d a ( 1. Sc ool of M at ematical and Computer Science, Nanjing Normal University , Nanjing 210097, C ina; 2. Department of Automation, Tsing ua Universit y, Beijing 100084, C ina. Correspondent: XU Jian- ua, E -mail: @ . . . ) xujian ua email njnu edu cn Abstract: Vapnik and is collaborators proposed a useful algorit m : support vector mac ines, w ic can implement t e structural risk minimizat ion principle in statistical learning t eory . T is novel algorit m andles t e classification problems successfully. Since t en more attentions ave been paid to statistical learning t eory and support vector mac ines. T e attractive researc areas are to improve or modify support vect or mac ines by optimizat ion tec niques, and to design t e novel non-linear mac ine learning algorit ms based on statistical learning t eory and some ideas in support vector mac ines, etc. T e advances in suc algorit m studies in t e last ten years are reviewed. Key words : mac ine learning; stat istical learning t eory; support vector mac ines 1引　　言 ) , , Vapnik