For support vector machine(SVM),when training set is very large,especially when there are many support vectors,the process of learning requires a great deal of EMS memory, and the speed of count is very slow.In this paper,a new fast classification algorithm is presented to train SVM by selecting border vectors which may be the support vectors,so as to reduce training samples and to increase training speed.Experiment results show that the algorithm not only acquires the same precision with that of the classi...