论文部分内容阅读
本文提出的模糊神经网络系统(FNNS)由三部分组成:输入信息的模糊化处理,模糊神经网络,模糊决策。采用神经计算的方法来获取模糊推理规则,研究模糊一致矩阵动态确定神经网络权向量的方法,解决目前神经网络学习过程中常常出现的训练样本不易确定、训练过程难于收敛等问题。在模糊决策方面,研究了最大关联隶属原则,解决了目前常用的最大隶属原则和模糊质心法所存在的丢失信息较多、不太适合离散论域等问题。
The proposed fuzzy neural network system (FNNS) consists of three parts: fuzzy information processing, fuzzy neural network, fuzzy decision-making. The method of neural computation is used to obtain the rules of fuzzy inference. The fuzzy consistent matrix is used to determine the weight vector of neural network dynamically to solve the problems that the training samples often appear in the neural network learning process are not easy to determine and the training process is difficult to converge. In the aspect of fuzzy decision-making, the principle of maximum association is researched, which solves the problems of the most commonly used principle of affiliation and the existence of missing information in fuzzy centroid method, which is not suitable for discrete domain.