自构型神经网络的设计

Design of Self Configuring Neural Network

运用人工神经网络的ＢＰ理论，设计并构造了一种自适应变步长自构型神经网络，通过大量的数据分析及网络训练，给出了权值修正的具体公式及变步长与误差超曲面形状的关系，提出了变步长加快学习收敛速度的一种直观解释．

An adaptive self configuring nueral network with varying step is designed and constructed by using BP theory of artificial nueral network.Through a vast amount of data analysis and network trains,a weigh value correcting formula is presented as well as the relations between the varying step and the configuration of supra curved plan of errors.An audio visual explanation of varying step is also proposed in order to quicken the learning of convergent speed. [