论文部分内容阅读
应用Δ目标函数法，研究了控制系统神经网络结构优化问题．给出一个推广算法及一个前向全连网络，并对用网络实现非线性映射进行仿真研究．结果表明：用推广算法训练前向全连网络，能同时得到优化结构与权值，训练速度快，适合控制系统神经网络使用
The Δ objective function method is applied to study the structural optimization of neural network in control system. A generalization algorithm and a forward fully connected network are given, and the simulation of the nonlinear mapping using the network is given. The results show that: using the generalized algorithm to train the forward all-connected network, the optimal structure and weight can be obtained at the same time, and the training speed is fast. It is suitable for the control system neural network