摘要
本文将模糊运算引入常规的Ｐｉ－ｓｉｇｍａ神经网络，大大增强了神经网络处理非线性和不确定性映射的能力．通过对Ｚａｄｅｈ模糊算子的等价变换，使模糊运算成为连续可微的函数，从而实现用梯度法对网络权值的更新．本文的结果显示，这种混合型神经网络在非线性建模和模糊建模等方面有重要的应用价值．
This paper develops a fuzzy neural network structure,namely fuzzy Pi--sigma neural network.The neural network is proved to have stronger nonlinear and uncertain mapping abilities than the conventional BP network and have significant application prospects in nonlinear modeling,fuzzy identification and self--organizing fuzzy control for complex systems.
基金
国家自然科学基金
关键词
神经网络
模糊集理论
非线性拟合
模糊建模
Pi--sigma neural network
fuzzy set theory
nonlinear approximation lfuzzy modeling
