具有正态模糊参数的模糊神经元网络学习机制  

A Learning Algorithm of the Neural Network with Bell-Shaped Fuzzy Parameters

作　　者：阿孜古丽.牙会浦 左孝凌[1] 

机构地区：[1]上海交通大学计算机系

出　　处：《云南大学学报（自然科学版）》1997年第S2期77-84,共8页Journal of Yunnan University(Natural Sciences Edition)

摘　　要：描述一个多层前馈式模糊神经元网络的学习机制．首先给出网络结构，其连接权值和阀值均为正态模糊数；然后采用模糊数学中区间数运算规则，对传统的ＢＰ算法进行了扩展，提出多层前馈式模糊神经元网络的基于正态模糊数的学习算法，用该算法训练后神经元网络能够完成确定或不确定的信息输入到近似结果输出的非线形映射；Described a learning algorithm of a multilayer feedforword fuzzy nerual network.First we gve the network architecture whose weights and thresholds are given as bell-shaped fuzzy numbers.Second we adopted the intervals algorithm of the fuzzy sets,extend the traditional BP-algorithm and proposed a bell-shaped fuzzy number based learning algorithm of a multilayer feedforword fuzzy neural network.After training with this proposed algorithm,the fuzzy neural network can perform the non-line ampping between definite or indefinite inputs vectors and crisp outputs.Last by learning the fuzzy reasoning rules we examine and analyze the ability of the proposed fuzzy neural network and its learning algorithm.

关 键 词：模糊神经元网络 BP算法 区间数 模糊推理 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]