B样条隶属函数的模糊神经网络研究

Research on Neural Fuzzy Network Using B Spline Membership function

把模糊子集的隶属函数取为Ｂ样条函数，利用Ｂ样条函数的正定性，紧密性和归一性，可以使得通常模糊系统输出中的分母削去，并且训练过程中权值的调整也限制在局部范围内。最后证明了其全局逼近特性。

B spline function as fuzzy subset's membership function is used. In terms of its positivity, compact support and normalization, the denominator of the defuzzified output is diminished, and the adjustments of the weights are confined locally. Finally, the property of the universal approximation is proved.