加快SMO算法训练速度的策略研究  

Research on improving speed of SMO

作　　者：骆世广[1] 骆昌日[2] 

机构地区：[1]广东金融学院应用数学系,广州510521 [2]华中师范大学网络学院,武汉430079

出　　处：《计算机工程与应用》2007年第33期184-187,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.10471045; No.60433020);广东省自然科学基金(the Natural Science Foundation of Guangdong Province of China under Grant No.970472; No.000463; No.04020079);广东科技公关计划(the Key Technologies R&D Program of Guangdong Province; China under Grant No.2005B10101010);霍英东基金( No.91005);教育部人文社科基金(No.2005-241);广州市天河区科技攻关项目(No.051G041);华南理工大学自然科学基金(No.B13-E5050190)

摘　　要：SMO(序贯最小优化算法)算法是目前解决支持向量机训练问题的一种十分有效的方法,但是当面对大样本数据时,SMO训练速度比较慢。考虑到在SVM的优化过程中并不是所有样本都能影响优化进展,提出了两种删除样本的策略:一种是基于距离,一种是基于拉格朗日乘子的值。在几个著名的数据集的试验结果表明,两种策略都可以大大缩短SMO的训练时间,特别适用于大样本数据。SMO （Sequential Minimal Optimization） algorithm is a very efficient method for training SVM.However,the training speed of SMO is very slow for the large-scale datasets.As only parts of the samples may affect the optimization of SVM,two strategies have been put forward for deleting the samples which are based on the distance and the value of Lagrange multiplies respectively.Experiments on several benchmark datasets have been done and the results show that the training time of the two strategies is reduced greatly,especially for the large-scale problems.

关 键 词：支持向量机 序贯最小优化算法 SHRINKING 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]