基于遗传算法的BP神经网络在高程拟合中的应用

Application of BP neural network based on GA in height fitting

针对传统BP神经网络难以选取具有全局性的初始点的缺点,使用改进的遗传算法全局搜索优化神经网络各层之间的连接权和阈值,提高了BP神经网络的收敛速度和泛化能力。结合高程拟合算例进行训练检验,证明该方法是一种改进BP神经网络的有效方法。

Because the weakness of traditional BP neural network is difficuhy to select the initial point, the improved genetic algorithm with global searching is used for optimization of the link weight and the threshold of the neural network layers for improving the capability of traditional BP neural network. By comparison, the convergence rate and generalization ability of BP based on genetic algorithm are higher than that of the traditional BP neural network. In the example of height fitting, it is proved that the improved algorithm is efficient.