摘要
从函数逼近理论出发，用一组正交基函数作为三层前向神经网络各隐含单元的输出特性，以其加权和作为网络的非线性输出，构成一种单输入单输出正交基函数神经网络模型．建立了多输入的多项式基函数神经网络，并给出了非线性静态特性拟合，ＸＯＲ特性和动态特性拟合计算机仿真结果．
Based on the function approximation theory, this paper presents a threelayer forward neural network firstly. It is an orthonormal basis SISO neural network model. The neurons' output character in the middle layer is an orthogonal system of polynomial, and the weighted sum of their outputs forms the output of the neural network. Next, we create a polynomial basis multi-input neural network. Finally, the simulation results of non-linear static, dynamic and XOR approximating are given.
出处
《湖南大学学报（自然科学版）》
EI
CAS
CSCD
1996年第2期84-89,共6页
Journal of Hunan University:Natural Sciences
基金
国家教委博士点基金
关键词
神经网络
正交基函数
模型辨识
neural network
orthonormal basis function
identification
