A formal analysis of stopping criteria of decomposition methods for support vector machines

被引:276
作者
Lin, CJ [1 ]
机构
[1] Natl Taiwan Univ, Dept Comp Sci & Informat Engn, Taipei 106, Taiwan
来源
IEEE TRANSACTIONS ON NEURAL NETWORKS | 2002年 / 13卷 / 05期
关键词
asymptotic convergence; decomposition methods; stopping criteria; support vector machines (SVMs);
D O I
10.1109/TNN.2002.1031937
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In a previous paper, we proved the convergence of a commonly used decomposition method for support vector machines (SVMs). However, there is no theoretical justification about its stopping criterion, which is based on the gap of the violation of the optimality condition. It is essential to have the gap asymptotically approach zero, so we are sure that existing implementations stop in a finite number of iterations after reaching a specified tolerance. Here, we prove this result and illustrate it by two extensions: nu-SVM and a multiclass SVM by Crammer and,Singer. A further result shows that, in final iterations of the decomposition method, only a particular set of variables are still being modified. This supports the use of the shrinking and caching techniques in some existing implementations. Finally, we prove the asymptotic convergence of a decomposition method for this multiclass SVM. Discussions on the difference between this convergence proof and the one in another paper by Lin are also included.
引用
收藏
页码:1045 / 1052
页数:8
相关论文
共 13 条
[1]   Training ν-support vector classifiers:: Theory and algorithms [J].
Chang, CC ;
Lin, CJ .
NEURAL COMPUTATION, 2001, 13 (09) :2119-2147
[2]   LIBSVM: A Library for Support Vector Machines [J].
Chang, Chih-Chung ;
Lin, Chih-Jen .
ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY, 2011, 2 (03)
[3]  
CORTES C, 1995, MACH LEARN, V20, P273, DOI 10.1023/A:1022627411411
[4]  
CRAMMER K, 2001, J MACHINE LEARNING R, V2, P265
[5]   A comparison of methods for multiclass support vector machines [J].
Hsu, CW ;
Lin, CJ .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2002, 13 (02) :415-425
[6]  
Joachims T, 1999, ADVANCES IN KERNEL METHODS, P169
[7]   Convergence of a generalized SMO algorithm for SVM classifier design [J].
Keerthi, SS ;
Gilbert, EG .
MACHINE LEARNING, 2002, 46 (1-3) :351-360
[8]   Asymptotic convergence of an SMO algorithm without any assumptions [J].
Lin, CJ .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2002, 13 (01) :248-250
[9]   On the convergence of the decomposition method for support vector machines [J].
Lin, CJ .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2001, 12 (06) :1288-1298
[10]  
OSUNA E, 1997, PROC CVPR IEEE, P130, DOI DOI 10.1109/CVPR.1997.609310