基于RBF神经网络的射频功放器件大信号建模方法研究

基于多层前向神经网络对任意非线性连续函数有较好的逼近效果，对BP(反向传播算法)神经网络和RBF(径向基函数)神经网络作了理论上的分析比较。并采用实际数据进行训练。说明了RBF神经网络在逼近精度和速度上都要优于BP神经网络。最后，以RBF神经网络作为函数逼近器对射频功率器件建立了大信号特征模型，并进行了模型检验，证明了基于RBF网络的建模方法具有较高的精度。

Large-signal modeling method of RF power amplifier based on RBFNN

As multilayer feedforward neural network can approximate any continuous function accurately, the authors compare RBF neural network with BP neural network in theory and use actual data to prove that RBF neural network is greater than BP neural network in approximation accuracy and rapidity. At last, the authors use RBF neural network as function approximator to build large-signal model of RF-power component and verify that the model based in RBF neural network can accurately simulate the behavior of RF-power component.