论文部分内容阅读
在 GGAP-RBF 算法的基础上,提出 RBF 神经网络的一种改进算法,结合网络中隐层神经元径向基函数的宽度自适应调整算法和重合度阈值的动态调整方法.通过函数逼近领域中的3个 Benchmark 问题,改进算法与RAN、RANEKF、MRAN、IRAN 和 GGAP-RBF(GAP-RBF)算法做比较.仿真结果表明在需要较少隐层神经元和训练时间前提下,改进算法训练的网络有较好的泛化能力.
Based on the GGAP-RBF algorithm, an improved RBF neural network algorithm is proposed, which combines the width adaptive adjustment algorithm of the hidden layer neuron radial basis function and the dynamic adjustment method of the coincidence threshold. By using the function approximation , The improved algorithm is compared with the algorithms of RAN, RANEKF, MRAN, IRAN and GAP-RBF (GAP-RBF) .The simulation results show that the proposed algorithm can improve the performance of training algorithm with fewer hidden neurons and training time Network has better generalization ability.