一种改进的快速支持向量机分类算法研究

快速的支持向量机分类算法--FCSVM对支持向量集采用变换的方式,用支持向量集的子集代替全部支持向量进行分类计算,在保证不损失分类精度的前提下使得分类速度较传统SVM算法有较大提高. 为了获得最小的支持向量子集,同时避免支持向量的移动,对FCSVM算法进行了改进.采用二分法优化分类函数中的支持向量数,给出了变换矩阵存在的充要条件及构造方法,减少了计算量.实验结果表明,改进的快速分类算法较大幅度地减少了计算复杂度,提高了分类速度,尤其在训练集规模庞大、支持向量数量较多的情况下,效果更加明显.

Research on an improved fast classification algorithm of support vector machines

The fast classification for support vector machines(FCSVM) performs transformation on the full set of support vectors,a subset of support vectors,which contains fewer support vectors,is used in classification.The speed of classification is much faster than that of conventional SVM under the condition that the precision of classification does not decline.In order to obtain minimal subset of support vectors and avoid movement of support vectors,the improvement on the method is presented.The minimal subset of support vectors is gotten by dichotomy.The condition of sufficient and necessary transform matrix existing and its construction method are given,the amount of computation is reduced.The experimental results show that the improved algorithm can remarkably reduce the computation complexity and the speed of classification is the fastest,especially in the case of large number of support vectors.