摘要
模糊规则的提取和隶属度函数的学习是模糊系统设计中重要而困难的问题。自适应神经网络模糊推理系统(ANFIS)能基于数据建模,无须专家经验,自动产生模糊规则和调整隶属度函数。在建立一个初始系统进行训练时,其隶属度函数的类型、隶属度函数的数目以及训练次数都是待定的,这三个参数的选择直接影响系统训练后的效果,它们的确定方法有待研究。该文应用自适应神经网络模糊推理系统的方法对一个典型系统进行建模仿真,并阐述这三个参数的寻优方法。
Extraction of fuzzy rules and learning of parameters of membership functions are vital but difficult when designing a fuzzy system. Applying Adaptive Neural - Fuzzy Inference System （ANFIS） can produce fuzzy rules and adjust membership functions automatically based on data without experience of experts. When setting up an initialized system to train, the type of membership functions, the number of membership functions and the time of training are all variables, and the choice of these parameters will directly affect the result of modeling, but the method for ensuring these parameters still needs research. This paper gives the simulation example for modeling a typical system with Adaptive Neural - Fuzzy Inference System and expatiates the method for choosing these three parameters.
出处
《计算机仿真》
CSCD
2005年第8期140-143,共4页
Computer Simulation
关键词
自适应神经网络
模糊系统
隶属度函数
Adaptive neural network
Fuzzy system
Membership functions
