用于实时识别时域信号的神经网络统计模糊算法

Statistic Fuzzy Algorithm for Neural Networks in Real-time Signal Recognition

本文讨论了一种基于统计训练和模糊判断的非监督神经网络算法，网络的权重是在已知样本的概率分布下得到的，输出单元兴奋状态及兴奋程度由被识别样本与确定权重的内积和被识别样本的模的隶属函数确定，网络的识别采用模糊聚类分析方法。最后给出了计算机模拟结果，证实了方法的正确性和可靠性。

A unsupervised neural networks algorithm based on statistic train and fuzzy decision is presented All weights of the neural networks can be determined under known sample's possibility distribution Active state and degree of the output nets are defined by the inner products between identified samples and obtained weights,and fuzzy function of the moduloes of identified samples. Fuzzy classification method is used to neural networks identification process. Finally, feasibility and practical applicability of this method has been confirmed by computer simulations.