摘要
基于神经网络法处理非线性问题的优势,探讨了神经网络法用于地铁隧道变形预测中影响网络收敛的各技术参数的选择问题,提出了防止网络训练过拟合及局部最小的方法。
On the basis of the advantage of the neural network method for processing nonlinear problem, some technological problems such as the election of technic parameteres may affect the network convergene in the prediction of tunnel deformation are analyzed. The method preventing exceed training and local optimization is presented.
出处
《大地测量与地球动力学》
CSCD
北大核心
2007年第1期80-84,共5页
Journal of Geodesy and Geodynamics
关键词
地铁隧道
变形预报
神经网络
归一化
反向传播算法
subway tunnel, deformation prediction, neural network, unitary, back-propagation
