支持向量机SVM的研究进展.pdfVIP

支持向量机( SVM )的研究进展 7 支持向量机( SVM ) 的研究进展 , , ( , 100022) :介绍了基于分解思想的支持向量机的训练算法, 主要有两大类: 块算法和固定工作样本集算法及 其对应的快速算法 SVMlightSMOBSVM 和L IBSVM;介绍了支持向量机的变形算法多类分类算法及 模型选择问题; 并且对近年来支持向量机在实际中的应用做了简要介绍, 最后讨论了支持向量机亟待解 决的问题并对其发展进行了展望 :支持向量机; 二次型规划问题; 多类支持向量机; 模型选择 : T 391 : A : 1000- 8829( 2006) 05- 0007- 06 Som e D evelopm en ts on Support Vector M ach in e LI Xiaoyu, ZHANG Xinfeng, SHEN Lansun (Signal Inform ation rocessing Laboratory, Beijing Un iversity of Technology, Beijing 100022, Ch ina) Abs tract: The training algorithm based on decomposition algorithm is introduced, which can be classified into two categories: chunking algorithm, fixed training set algorithm and its corresponding fast algorithm: SVMligh,t sequentialm inmi al optmi ization algorithm, BSVM and LIBSVM. The transformative algorithm based on SVM, multiclass SVM andmodel selection are also presented. Then, the applications of support vectormachine are in troduced. Lastly, the problems in SVM to be solved are discussed, and itsdevelopment orientation is pointed ou.t K ey w ord s: support vectormachine; quadratic problem; multiclass SVM; model selection ( SVM, support vector machine) V apnik CortesV apnik 1995, 20 90 Vapnik Chervonenkis [ 1] [ 2] , ( VC VC, , ), VC, , ( SRM, , structural risk m inmi ization), R (w ) R em p (w ) + ( n ) ( 1) , h SVM ; , , ( generalization) , n , VC(h ), , ,