摘要
论文介绍了一种年轻的机器学习方法——支持向量机,详细论述了目前主要的支持向量机的训练算法,包括:二次规划算法,分解算法和增量算法。通过实验验证了普通二次规划算法的缺陷,比较了三种典型的SVM分解训练算法的性能,说明了其相对于二次规划算法的优点和对SVM训练问题的适用性,指出了训练速度优劣的原因。最后指出了未来支持向量机训练算法研究的方向。
This paper introduces a new machine learning algorithm--Support Vector Machine.This paper presents main current SVM training algorithms in detail,including quadratic optimization algorithm,decomposition algotithm and incremental algorithm.The experimental results show the shortcomings of quadratic optimization algorithm.With the comparison of the performance of three classic SVM training algorithms,the authors have proved the advantages and applicability of them for SVM training problem over quadratic optimization algorithm,and have analysed the reasons causing the difference of training speed.In the end they have pointed out the research direction of SVM training algorithms in the future.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第17期79-81,84,共4页
Computer Engineering and Applications
关键词
支持向量机
训练算法
support vector machine,training algorithm
