一种基于GA和梯度下降法的快速模糊系统学习方法  

A Fast Learning Method of Fuzzy Systems Based on GA and Gradient Descent Method

作　　者：武妍[1] 施鸿宝[1] 

机构地区：[1]上海铁道大学计算技术研究所

出　　处：《上海铁道大学学报》1999年第6期16-20,26,共6页

基　　金：上海市教委重点学科项目

摘　　要：分析了现有模糊系统的两类主要的学习算法存在的缺陷。针对半梯形和三角形隶属度函数，提出了一种保证隶属度函数ε－完备性的方法。实现了一种新的基于遗传算法和梯度下降方法的快速模糊系统学习算法。通过实例进行了模拟，验证了该方法的高效性，以及保证隶属度函数完备性和模糊集合语义一致性的优点。This paper first makes an intensive analysis of two kinds of learning methods for fuzzy systems using genetic algorithm and gradient method. A method that guarantees the ε -completeness of membership functions and consistence of fuzzy sets semantics is proposed.Moreover,a new fast learning method of fuzzy systems both based on genetic algorithms and gradient method is proposed.Some experiments are also made and one simulation result is presented to show the high effectiveness and some other advantages of guaranteeing the completeness of membership functions and consistence of fuzzy sets semantics.

关 键 词：模糊系统 算法 遗传算法 梯度下降方法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]