摘要
提出一种迭代再权q范数正则化最小二乘支持向量机(LS SVM)分类算法。该算法通过交叉校验过程选择正则化范数的阶次q(0<q<∞),具有分类效果稳定、收敛快、运行速度快的特点。使用该算法和LS SVM对比3组癌症数据,实验结果表明,该算法能够实现自适应特征选择,且比LS SVM推广能力强,在算法耗时方面优于LS SVM。
This paper proposes the classification algorithm of fast iterative reweighted q-norm regularization Least Squares Support Vector Machine（LS SVM）. The proposed algorithm can select q value via cross-validation, where O〈q〈oc,, and has the characteristic of stability, quick-converging and low time complexity. In order to test the efficiency of the proposed algorithm, it is applied to three cancer datasets. Experimental results show that the presented algorithm can obtain adaptively feature selection with better generalization performance for the classification problems than LS SVM. and its training speed is much faster than LS SVM.
出处
《计算机工程》
CAS
CSCD
2012年第3期166-168,共3页
Computer Engineering
关键词
迭代再权方法
q范数
最小二乘支持向量机
正则化
特征选择
分类算法
iterative reweighted method
q-norm
Least Squares Support Vector Machine（LS SVM）
regularization
feature selection
classification algorithm
