Optimization research of genetic neural network based on Scilab

摘要

Radial basis function (RBF) network is one of the significant neural networks. It has been used successfully in various fields. But in RBF network approximation algorithm, the initial value of the network weights, Gauss function center vector and broad-based vector is not easy to determine, and when these parameter choice is undeserved, RBF network approximation precision will decline and even the serious consequences of network spread will be produced. By using genetic algorithm in this paper, which can better realize RBF network parameter optimization, thereby increasing the accuracy of approximation. Scilab is open source software and has good simulation capabilities. Experiments using Scilab shows that the optimization method of genetic neural network is feasible and results are satisfied.