论文部分内容阅读
提出了一种新型的神经网络的训练方法，包括输入层节点的自动寻找技术和隐含层节点的自动生成技术．为了验证该算法的正确性，利用该算法对同向回转观机驱动振动系统进行了神经网络逼近分析，计算机模拟结果表明该算法可以准确地确定输入层和隐含层节点的数目．
A new neural network training method is proposed, which includes the automatic search of input nodes and the automatic generation of hidden nodes. In order to verify the correctness of the algorithm, this algorithm is used to approximate the neural network of the homotopy oscillating system. The computer simulation results show that the algorithm can accurately determine the number of input layer and hidden layer nodes.