节点文献
应用自递归神经网络(SRNN)预测结构响应
PREDICTION OF STRUCTURAL RESPONSE BY SELF-RECURRENT NEURAL NETWORK
【摘要】 本文提出了一种新的自递归神经网络结构.这种网络结构由全递归网络改造而成,只有一个隐层,而且隐单元仅存在自递归.研究了这种网络的学习算法.为了保证快速学习收敛,应用Lyapunov函数得到一种自适应学习率方法.用这种方法对一两层建筑结构响应进行在线预测.计算机仿真结果表明,这种网络学习算法是有效的,并且是可行的
【Abstract】 A new neural paradigm called Self-Recurrent Neural Network (SRNN) is presented here. The architecture of SRNN is a modified model of the fully connected recurrent neural network with one hidden layer, and the hidden layer is comprised of self-recurrent neurons. A generalized dynamic back-propagation algorithm (DBF) is established. To guarantee convergence and faster learning, an approach using adaptive learning rates is developed by introducing a Lyapunov function. Convergence theorem for the adaptive back-propagation algorithm is developed for on-line prediction of the response of a two-story building excited by external force. Results form computer simulation studies demonstrate that the new DBP is valid and feasible in on-line predicting structural response.
【Key words】 self-recurrent neural network; learning rate; structural response;
- 【文献出处】 土木工程学报 ,China Civil Engineering Journal , 编辑部邮箱 ,1998年02期
- 【分类号】TU311.41
- 【被引频次】42
- 【下载频次】124