摘要
本文提出一种神经模糊系统模型，其中模糊规则前件用л隶属函数（形状类似于三角形隶属函数，但具有平滑性）表达，给出了类似于BP的参数学习算法．对于平滑函数近似问题的仿真结果表明，与模糊规则前件使用三角形隶属函数的神经模糊系统模型相比，本文提出的模型具有学习过程更加稳定平滑和逼近误差小的优点．对这两种模型性能上的差异做了定性解释．
A neural fuzzy system model where the antecedents of fuzzy rules are expressed bymembership functions (similar to trangular membership functions but smooth )is proposed and itsparameter learning algorithm like the BP is derived in this paper Simulation results for smoothfunction approximation problem indicate that the learning processes are more stable and smooth andthe approximation errors are more small for the proposed model compared to the model using thetriangular membership functions Performace differences between the two models are explainedqualitatively
出处
《计算机学报》
EI
CSCD
北大核心
1998年第S1期121-126,共6页
Chinese Journal of Computers
关键词
神经模糊系统
隶属函数
学习算法
平滑
Neural fuzzy system membership function learning algorithm smooth
