SVM Algorithm Based On Boundary Vector

In order to improve the training speed of SVM,the method of reducing the size of training set is proposed to solve the problem of high computational complexity of traditional support vector machine.In view of the traditional support vector machine in the presence of raw data is too large lead to the training speed is too slow,and affect the performance of support vector machine(SVM)the characteristics of the support vector is often located in the border,put forward two kinds of boundary vectors of support vector machine(SVM)algorithm.One is based on the boundary of the center vector vector extraction algorithm,through the calculation of two kinds of center class center of vertical split surface,according to the distance of each sample to the vertical split surface to retain samples;The other one is based on vector projection extraction algorithm,all sample are projected to the center of the two kinds of attachment,according to the size of the distance to retain,and eliminate noise samples in the retention samples.Calculate the distance of each sample to the center of the two classes,according to the size of the distance to retain the samples,to eliminate the noise samples in the samples.Numerical experiments show that the improved algorithms can ensure the classification performance of support vector machine(SVM),effectively improve the classification efficiency of support vector machine(SVM).