Application of Mathematical Statistical Analysis in Neural Network Model Optimization

摘要

In order to solve the problem of uneven distribution of initial population caused by symbol coding in genetic neural network, this paper starts from the working principle of neural network and analyzes the influence of excitation function and system parameters on the output results by using mathematical statistics method. Then a two-stage combinatorial genetic algorithm optimization model is proposed to find the optimal combination of S-type function and RBF function in the network, so that the unconventional nodes can be adjusted in the process of back propagation. Through the simulation test, it is proved that the scheme makes full use of the nonlinear characteristics of various excitation functions, which is conducive to improving the training speed of BP neural network and the convergence speed and prediction accuracy of the model.