Research on the Classification Method Based on BP Neural Network and SVM

This paper introduces the classification principle of BP neural network and SVM algorithm.Additional momentum factor and stochastic gradient descent algorithm is an important method of BP neural network optimization.Google Labs MNIST handwritten digital library are used to study the momentum factor and a random number and impact of different kernel functions.SVM classification performance provides practical application of the model provide a basis for selection.Also the author studies the performance of the two algorithms under different number of samples.Experiments show that there are fewer than the number of samples SVM BP has higher generalization ability.Finally,the characteristics of the two algorithms are given as a hierarchical classification and future research directions.