An Image Classification Scheme for Improved Convolutional Neural Networks

In order to solve the defects of low classification accuracy and slow convergence speed in the traditional image classification task, this paper designs an improved neural network based on particle swarm algorithm. The network improves the convolutional neural network by changing the method of bias value update during back propagation. The traditional method of changing the bias value is the gradient descent method, but there may be the problem of gradient disappearance and gradient exploding. In this paper, the particle swarm algorithm is used instead of the gradient descent method to solve the problem of gradient disappearance. It is analyzed by the test with handwritten data set. The convergence speed and accuracy of the method are also improved compared with the traditional convolutional neural network.