Application of BP neural network based on GA in function fitting

To avoid BP algorithm´s shortcoming of falling into local minima and to take advantage of the genetic algorithm´s globe optimal searching, combining genetic algorithm with BP neural network, we established the model of nonlinear function approximation based on genetic - algorithm - optimized BP neural network, analysed the toplogical structures of networks and described its learning algorithm. In the model, the initial weights and thresholds of BP network were optimized using genetic algorithm, and revised according to the negative gradient direction, the network was trained to get the optimal solution. The paper used BP network and genetic - algorithm - optimized BP network respectively to approximate the same nonlinear function. The simulation results show that the genetic - algorithm - optimized BP network has better nonlinear fitting ability and prediction accuracy.