摘要
基于卡尔曼滤波的BP神经网络方法,对大坝变形观测数据进行滤波处理,用滤波后的数据参与BP网络的训练,使网络具有动态特性,减小了神经网络陷入局部极小值的可能性,提高了神经网络的泛化能力。实例证明,该方法具有很高的预测精度和较强的泛化能力。
A new dam deformation perdition model of BP neural network based on Kalman filtering are put forward. The filtered sample data is used for BP training, it makes the network have dynamic properties and reduces the possibility of the local minimum value of Neural network. The precision and generalization ability of BP based on Kalman filtering are higher than those of the traditional BP neural network. Through the example it is proved that the new algorithm is of high accuracy and generalization ability in the data processing.
出处
《大地测量与地球动力学》
CSCD
北大核心
2009年第4期124-126,共3页
Journal of Geodesy and Geodynamics
基金
国家自然科学基金(40672173)
