论文部分内容阅读
1 引言模糊系统建模一般将经过系统结构辨识和系统参数估计两个阶段。在辨识阶段,主要决定输入变量及其相互关系、模糊规则数、输入输出空间划分和系统参数的初值;在估计阶段,主要用来调整系统参数以使得系统的输出与目标输出的差值尽可能小。对于系统参数估计阶段的参数调整,人们已提出一些自动方法。对于系统结构辨识阶段,也产生了如模板法、聚类法和决策树法等,但这些方法一般都需要人工干预。其中模糊规则的生成与调整以及隶属度函数的选取是系统结构辨识阶段的主要问题,文提出了用神经网络自动生成模糊规则并进行隶属度形状调整,从而构成模糊神经网络。Wang提出自动分割输入空间的方法,Lin提出三阶段学习算法的模糊神经网络。
1 Introduction Fuzzy system modeling will generally be two phases of system structure identification and system parameter estimation. In the stage of identification, the input variables and their relations, the number of fuzzy rules, the input and output space and the initial values of the system parameters are mainly decided. In the estimation stage, the system parameters are mainly used to adjust the system parameters such that the difference between the system output and the target output May be small. For the adjustment of the parameters of the system parameter estimation stage, some automatic methods have been proposed. For the stage of system structure identification, such as template method, clustering method and decision tree method also emerge, but these methods generally require human intervention. The generation and adjustment of fuzzy rules and the selection of membership functions are the main problems in the phase of system structure identification. In this paper, the fuzzy neural network is constructed by automatically generating fuzzy rules and adjusting the shape of membership degree. Wang proposed a method of automatically segmenting the input space. Lin proposed a three-stage learning algorithm for fuzzy neural networks.