基于多种正交基函数的模块化过程神经元网络  

Modular Process Neural Network Based on Multi-Basis-functions

作　　者：邓韧[1] 李著信[1] 樊友洪[1] 

机构地区：[1]后勤工程学院,重庆400016

出　　处：《计算机工程与应用》2005年第30期69-71,74,共4页Computer Engineering and Applications

摘　　要：提出一类基于多种正交基函数的模块化过程神经元网络模型,它融入了多时变输入的空间聚合和作用域限制的时间累积,并采用多种正交基函数在较小网络规模的条件下保证系统各种输入输出的精度,应用混合隐含层综合考虑了系统多类型输入对系统的作用,并应用模块化级联的方式在一定程度上减小了网络的总体容量,从而提高了整个网络的学习速度。应用实例的训练及仿真结果证明了该模型的可行性和有效性。A modular process neural network model based on multi-basis-functions is brought forward in this article,in which the spatial aggregation and temporal limited accumulation of discrete-time inputs are involved.Under the small scale of network,multi-basis-functions are used to ensure the precision of inputs and outputs of system.All kinds of inputs play their roles in system by Blending Hidden-layer,and the gross capacity of network is reduced by modular cascade,meanwhile,the convergence rate of learning is enhanced.The results of instance show the feasibility and effectiveness of the model.

关 键 词：过程神经元网络 模块 正交基函数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]