Research on BP Neural Network Optimization Based on Improved Global Artificial Fish Swarm Algorithms

Abstract

This paper uses the improved global artificial fish swarm algorithm to optimize BP neural network, which reduces the training error of the neural network. Firstly, the structure of BP neural network is established to determine the number of network layers and neurons. Secondly, the improved global artificial fish swarm algorithm is used to optimize the algorithm, which mainly searches for the initial weights and thresholds of the global optimum according to the convergence and search ability of the algorithm. Finally, the optimized BP neural network is used to detect, the weights and thresholds are assigned to the BP neural network. Then the weights and thresholds are trained to minimize the training error. Through experimental analysis, the improved global artificial fish swarm algorithm can optimize BP neural network better, reduce errors and training times, improve the training speed of the network, and enrich the research on the improvement of the neural network.