一种SVM增量训练淘汰算法  

A Removing Algorithm for Incremental SVM Training

作　　者：史朝辉[1] 王晓丹[1] 杨建勋[1] 

机构地区：[1]空军工程大学导弹学院,陕西三原713800

出　　处：《计算机工程与应用》2005年第23期187-189,212,共4页Computer Engineering and Applications

基　　金：陕西省自然科学基金资助项目

摘　　要：基于KKT条件分析了样本增加后支持向量集的变化情况,深入研究了支持向量分布特点,提出了一种新的支持向量机增量训练淘汰机制——挖心淘汰算法。该算法只需设定一个参数,即可对训练数据进行有效的遗忘淘汰。通过对标准数据集的实验结果表明,使用该方法进行增量训练在保证训练精度的同时,能有效地提高训练速度并降低存储空间的占用。This paper presents a novel approach to incremental Support Vector Machine （SVM） learning algorithm.It analyses the possible change of support vector set after new samples are added to training set.Furthermore,the properties of SV set are analyzed thoroughly.After that,a new improved incremental SVM learning algorithm is proposed,which is based on a sifting method.This algorithm utilizes distribution peculiarity of the SV set while the incremental training is proceeded,and thus makes it possible to discard samples optimally.The theoretical analysis and experimental results show that this algorithm can not only improve the training speed,but also reduce storage cost.

关 键 词：支持向量机 增量训练 淘汰机制 边界矢量 挖心淘汰算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]