Optimization of Neural Network based on Genetic Algorithm and BP

In order to improve the intelligence, high efficiency, humanization of the type of the search and eliminate games, and also to improve the search performance and rule out the accuracy of the target during intelligent games running. This paper puts forward a comprehensive method that combines Genetic Algorithm, Neural Network and Back Propagation (BP) to solve the insufficiency of computing power and low efficiency by using a single algorithm in Intelligence games. In this method, Genetic Algorithm will be used in weight training of Neural Network first of all. It will not stop iterating until Genetic Algorithm evolves into a certain degree or network errors satisfies the requirements, and delivers the best chromosome we get to Neural Network. Then BP trains the data that runs through the Neural Network, which is Neural Network´s second training. Finally, the paper applies the new way in the Mine Clearance experiment. By comparing this experiment with only using Genetic Algorithm or Neural Network, it finds out that the proposed method significantly improves the minesweepers accuracy.