利用模糊感知器解决模糊可分样本的分类问题  

Classification of fuzzily separable training patterns based on fuzzy perceptron

作　　者：刘燕[1] 阎慧臻[1] 刘超[1] 

机构地区：[1]大连工业大学信息科学与工程学院,辽宁大连116034

出　　处：《大连工业大学学报》2009年第1期66-68,共3页Journal of Dalian Polytechnic University

摘　　要：针对一类用于解决分类问题的模糊感知器,提出完全随机输入的模糊δ-规则,并给出训练样本模糊可分的定义。实例表明,利用该算法可以有效地解决模糊可分样本的分类问题,在有限步迭代后就达到收敛,即有限步训练后网络能将所有样本正确分类。A fuzzy δ-rule training algorithm was proposed for a fuzzy perceptron, in which the training patterns were totally random inputted. A fuzzily separable definition for training patterns were presented. Numerical experiments showed that fuzzy perceptron based on fuzzy δ-rule training algorithm could effectively solve the classification problems for fuzzily separable training patterns, and the algorithm is finitely convergent, i. e. , the training patterns could be correctly classified by the net after fi- nite steps of iterating.

关 键 词：模糊感知器 模糊可分 分类 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]