带递归单元的模糊感知器的δ-规则的有限收敛性  

Finite convergence of δ-rule for a recurrent fuzzy perceptron

作　　者：刘燕[1] 阎慧臻[1] 刘超[1] 姜行健[2] 杨开兵[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 [2]大连理工大学机械工程与材料能源学部,辽宁大连16024

出　　处：《大连工业大学学报》2012年第5期376-378,共3页Journal of Dalian Polytechnic University

摘　　要：针对带递归的模糊感知器,提出模糊δ-规则,其中样本以完全随机顺序输入。证明了若训练样本模糊可分,在一定条件下,算法有限收敛,即有限步训练后网络能将所有样本正确分类,可以准确完成模糊可分样本的分类问题。A fuzzy δ-rule training algorithm was proposed for a recurrent fuzzy perceptron,in which the training patterns were supplied in completely stochastic order.It was proved that fuzzy δ-rule training algorithm was finitely convergent in the case that the training patterns were fuzzily separable.The training patterns could be correctly classified by the net after finite steps of iterating so that the perceptron could effectively solve classification problems.

关 键 词：递归 模糊感知器 模糊可分 

分 类 号：U495[交通运输工程—交通运输规划与管理]