摘要
提出一种基于模块化模糊子系统的分层模糊神经网络.该分层模糊神经网络基于高斯隶属函数,且功能上等价于一个TSK模糊系统.这种分层神经网络在保留了传统模糊神经网络很多优点的同时有效地抑制了“维数灾”问题,而且在模糊子系统中模糊规则的激活强度有所提高.仿真试验结果表明,该方法能获得更为简洁有效的模糊规则集.
A hierarchical fuzzy neural network based on module fuzzy subsystems （HM-FNNs） is proposed, which is built based on ellipsoidal basis function and is equivalent to a Takagi-Sugeno-Kang fuzzy system functionally. The HM-FNNs not only remains the full benefits of a traditional FNNs but also suppress the effects of the unwanted phenomenon, ＂the curse of dimensionality＂. It also offers one great advantage that all rule fire strengths are strong on average when passing through subsystem layers. The simulation results show that the proposed method can produce the compact and high performance fuzzy rule-base.
出处
《控制与决策》
EI
CSCD
北大核心
2006年第3期281-284,共4页
Control and Decision
关键词
聚类方法
分层模糊神经网络
进化规划
Clustering algorithm
Hierarchical fuzzy neural network
Evolutionary programming
