摘要
提出二阶反向传播神经网络的超松驰训练方法,证明了该算法的收敛性.将该网络及其新的训练方法用于非线性系统的自适应控制中,能够更有效、更快速地跟踪系统的参考输出.数值实验结果显示超松驰训练方法优于直接梯度优化算法,而且基于二阶反向传播神经网络的直接自适应控制效果更好.
A new training algorithm called BPSOR for second order Back Propagation Neural Network (BPNN) is proposed, and the convergence of this algorithm is proved. The use of this neural network and its new training algorithm in direct adaptive control for a class of Single-Input Single-Output systems can get more effective and faster track of the system's reference output. A numerical example is given to show that the BPSOR is better than direct gradient optimization method, and the second order BPNN-based direct adaptive control is better than the first order BPNN-based control.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第1期80-83,共4页
Pattern Recognition and Artificial Intelligence
关键词
二阶反向传播神经网络
权值训练
超松驰方法
收敛性
自适应控制
Second Order Backpropagation Neural Network, Weight Training, Successive Overrelaxation (SOR) Method, Convergence, Adaptive Control
