Abstract: When the number of support vectors or samples required to distinguish is large, its direct computation would take much time. In order to reduce the complexity of discriminant value computation, vector replacement method mainly for linear SVM and orthogonal vectors method mainly for nonlinear SVM are brought forward.