论文部分内容阅读
研究基于正态隶属函数的模糊神经网络的学习算法 .将模糊神经网络对一组样本的逼近误差表示为两组相互独立 ,可分批学习的可调参数的非负函数之和 .其中一组可调参数可通过令相应的非负函数为零直接求得 ,而与另一组可调参数相对应的非负函数就是用于这组参数学习的性能指标 .经对性能指标性质的分析给出了一种模糊神经网络的学习算法——二阶段变半径随机搜索法 .实例表明 ,这种方法简便易行 ,可使模糊神经网络达到较高的逼近精度
This paper studies the learning algorithm of fuzzy neural network based on normal membership function.The approximation error of a set of samples by fuzzy neural network is represented as the sum of nonnegative functions of two independent adjustable parameters that can be studied in batches, The adjustable parameters can be obtained directly by setting the corresponding non-negative function to zero, and the non-negative function corresponding to another set of adjustable parameters is the performance index used for the parameter learning of this group.Based on the analysis of the nature of the performance index A fuzzy neural network learning algorithm - two-stage variable radius random search method is given.Examples show that this method is simple and easy and can make the fuzzy neural network reach a higher approximation accuracy