Selection of initial weights and thresholds based on the Genetic Algorithm with the optimized Back-Propagation neural network

摘要

The Genetic Algorithm (GA) has been used for artificial neural network optimization of initial weights and thresholds matrix. The initial weights and thresholds which has been optimized is equivalent to set up a reference range for artificial neural network to search optimization, therefore the performance of improved artificial neural network could be enhanced. Matlab2010 has been used as a platform for experiments. With the actual case, the error between the algorithm of individual using the BP (Back-Propagation) network and the optimized GA-BP (Genetic Algorithm and Back-Propagation) neural network with the genetic algorithm has been carefully analyzed. Results show that all indicators improved for the latter.