A Convolutional Neural Network Model Based on Particle Swarm Optimization Algorithm

Nowadays, most users of convolutional neural networks set the parameters of the neural network architecture through experience, intuition, or repeated experimentation, which makes it difficult to effectively and quickly find the best convolutional neural network. To quickly and efficiently set the optimal parameters of the convolutional neural network, this paper uses the convolutional neural network as the basis for handwritten character recognition, and uses particle swarm optimization algorithm to optimize the internal structure of the convolutional neural network. Using the number and size of convolution kernels, dropout rate, learning rate, and hidden layer as parameters for particle swarm optimization algorithm to improve recognition accuracy. We used MNIST handwritten black and white digits as the dataset and optimized the parameters through PSO, resulting in a recognition rate of 99.56%. The experiment shows that this algorithm effectively improves the recognition rate of the model.