BP NEURAL NETWORKS STRUCTURE OPTIMIZATION BASED ON IMPROVED LMBP ALGORITHM

摘要

This paper presents an algorithm to optimize artificial neural networks structure based on constructive method. A LMBP algorithm is introduced about simplest BP neural network for function approximation. By rules of error changing based on quadratic error and gradient reducing, analyzing the optimization of the network's structure and adding hidden neurons or adding network layers one by one adaptively, as a result a proper structure of the network is got. Simulation experiments are provided to compare the approach with RAN algorithm for solving function approximation.