【摘要】 提出一种模糊神经网络 ( FNN)结构学习算法 ,根据输入样本动态构建 FNN的输入节点及其对应的输入隶属函数 ,从而实现动态确定 FNN的结构 ,大大减少了对初始学习样本数目的要求 ,提高了 FNN学习算法在实时控制中的适应能力 .仿真结果表明 ,这一算法很好地实现了对超出初始学习样本范围的其他样本的学习 .更多还原

【Abstract】 A dynamically learning method for fuzzy neural network structure was presented. The number of input fuzzy membership functions is adjusted by adding, pruning and combining the membership function knots according to the new samples. The parameters of membership functions are determined by the distance of samples. The results of simulation show that the proposed methods can learn in a satisfying accuracy with almost no requirement for the number or distribution of initial samples.更多还原