期刊文献+
任意字段
题名或关键词
题名
关键词
文摘
作者
第一作者
机构
刊名
分类号
参考文献
作者简介
基金资助
栏目信息

一种通用的基于梯度的SVM核参数选取算法 被引量：5

A Gradient Method for Choosing Kernel Parameters for SVM
原文传递
导出
摘要 核函数的选取是SVM分类器选取的核心问题.核函数的自动选取既可以提高分类器的性能,又可以减少人为的干预.因此如何自动选取核函数已经成为SVM的热点问题,但是这个问题并没有获得很好的解决.近年来对核函数参数的自动选取的研究,特别是对基于梯度的优化算法的研究取得了一定的进展.提出了一种基于梯度的核函数选取的通用算法,并进行了实验. It＇s well-known that the major task of the SVM lies in the selection of its kernel. Choosing the kernel parameters automatically could improve the performance of SVM while reduce the intervention of human user.However,the best way to determine the parameters of the kernel is still an open question.Recent advances in kernel-based learning algorithm have brought SVM closer to the goal of autonomy.This paper presents a uniform algorithm based on gradient for choosing the kernel parameters of the SVM.Experimental results show the feasibility of our approach.
作者 王永锋 黄思明 徐晶
机构地区 中国科学院科技政策与管理科学研究所 易方达基金管理有限公司
出处 《数学的实践与认识》 CSCD 北大核心 2011年第1期7-13,共7页 Mathematics in Practice and Theory
关键词 核函数 支持向量机 模型选取 kernel function SVM model selection
分类号 TP18 [自动化与计算机技术—控制理论与控制工程]
  • 相关文献

参考文献15

  • 1Joachims T. The maximum-margin approach to learningtext classifiers: method, theory and algorithms[D]. Ph D. Thesis,University of Dortmund, 2000.
  • 2Jaakola T S, Haussler D. Probabilitic kernel regression models[C]//D. Heckerman, J. Whittaker (Eds.), Proceedingsof the 1999 Conference on AI and Statistics, San Mateo, CA, Morgan Kaufmann, 1999.
  • 3Opper M, Winther O. Gaussian processes and svm: mean fieldand leave-one-out[M]. In: A.J. Smola, P. Bartlett, B. Scholkopf, D. Schuurmans (Eds.), Advances in Large Margin Classifiers, MIT press, 2000: 311-326.
  • 4Smola A. Learning with kernels[D]. Ph D Thesis, GMD First, Berlin, Germany, 1998.
  • 5Williamson R C, Smola A J, SchSlkopf B. Entropynumbers, operators and support vector kernels[M]. In: B.Scholkopf, C. Burges, A. Smola (Eds.), Advances inKernel Methods--Support Vector Learning, MIT Press, 1999: 127-144.
  • 6Vapnik V. Statistical learning theory[M]. New York: John Wiley &: Sons, 1998.
  • 7Vapnik V and Chapelle O. Bounds on error expectation for support vector machines[J]. Neural Computation, 2000, 12(9).
  • 8Chapelle O, Vapnik V. Choosing multiple parameters for support vector machines[J]. Machine Learn, 2002, 46 (1-3): 131-159.
  • 9Wahba G, Lin X, Gao F, Xiang D, Klein R, Klein B. The bias-variance trade-off and the randomized gacv, in:M. Kearns, S. Solla, D. Cohn (Eds.), Advances in Neural Information Processing Systems 11: Proceedings of the 1998 Conference, 11, MIT Press, Cambridge, MA, 1999, 620-626.
  • 10Bengio Y. Gradient-based optimization of hyper-parameters[J]. Neural Computation, 1999, 12(8).

同被引文献45

  • 1封筠,陈志军,李莉蓉.基于修正核函数的SVM分类器研究[J].系统仿真学报,2006,18(3):570-572. 被引量：10
  • 2Platt J C. Sequential minimal optimization: a fast algorithm for training support vector machines[J]. Advances in Kernel Methods Support Vector Learning, 1998, 208(4) : 1-21.
  • 3Scholkopf B, Smota A J. Learning with kernels support vector mahcines, regularization, optimization and beyond [M]. MIT Press, 2002: 187-200.
  • 4Ming H Q, Partha N, Yao Yuan. Mercer's theorem fea- ture maps and smoothing[J]. IEEE International Confer- ence on Tools with Artificial Intelligence (COLT), 2006, 40(5) : 154-168.
  • 5Tax D J, Robert P W. Support vector domain description [J]. Pattern Recognition Letters, 1999(20):1191-1199.
  • 6David M J Tax, Robert P W. Data domain description u- sing support veetors[C]//European Symposium on Artifi- cial Neural Networks, 1999 : 251- 256.
  • 7Juszczak P, Robert P W. Uncertainty sampling methods for one-class classifiers [J]. International Conference on Ma- chine Learning, 2003,9(12) :42-56.
  • 8Tax M J, Alexander Y P, Robert P W. Support vector data description applied to machine vibration analysis[C]//Sth Annual Conference of the Advanced Pattern Recognition, 1999,353:121-134.
  • 9王占全,徐慧.VisualC + +数字图像处理技术与工程案例_北京:人民邮电出版社,2009.
  • 10Serafeim Moustakidis, Giorgos Mallinis, Nikos Koutsias, et al. SVM-based fuzzy decision trees for classification of high spa-tial resolution remote sensing images I-J]. IEEE Transactions on Geoscience and Remote Sensing, 2012, 50 (1): 149-168.

引证文献5

二级引证文献16

数学的实践与认识
2011年 第1期

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部