论文部分内容阅读
针对径向基函数(RBF)神经网络构造时其结构和参数难以确定的问题,结合可拓理论对输入样本和基函数的中心向量建立物元模型,并借鉴第2类型可拓神经网络(ENN2)的聚类思想,根据样本分布,采用可拓分析及可拓变换动态调整隐节点数目和基函数中心,从而提出基于可拓理论的RBF(ERBF)神经网络.同时,通过UCI标准数据集进行了测试,并通过应用实例进行了验证,结果表明,ERBF结构和参数的确定方法简单、收敛速度快,且泛化精度、鲁棒性和稳定性均显著提高.
Aiming at the problem that the structure and parameters of radial basis function (RBF) neural network are difficult to be determined, the matter-element model of center vectors of input samples and basis functions is established based on extension theory, and the second type of RBNN (ENN2 ), The RBF (ERBF) neural network based on extension theory is proposed according to the distribution of samples, and the number of hidden nodes and the basis function center are dynamically adjusted by using extension analysis and extension transformation. At the same time, by the UCI standard data set The results show that the method of ERBF structure and parameter determination is simple, the convergence speed is fast, and the generalization accuracy, robustness and stability are all improved remarkably.