Optimization of Neural Network Based on Improved Genetic Algorithm

Neural network and genetic algorithm have attracted a great deal of attention as methods and theories realizing artificial intelligence recently. The combination of these two is drawing more and more attention. This paper demonstrates the possibility of combining neural network with genetic algorithm. An improved genetic algorithm for the learning of neural network´s connection weights is presented. According to the XOR problem, it has been indicated that the new method has the capability in fast learning of neural network and the capability in escaping local optima and initial weights. The algorithm gets performance far superior to traditional genetic algorithm and BP algorithm in all sides.