Nova Abordagem de Redes Neurais de Função de Base Radial de Valores Complexos para Modelagem Comportamental de Amplificadores de Potência

This work deals with the behavioral modeling of power amplifiers for wireless communication systems based on complex-valued radial basis function neural network (CVRBFNN). In this work, modifications on the CVRBFNN input and output are proposed with the purpose of optimizing its usage for low-pass equivalent modeling. In comparison with a previous approach of same computational complexity, results from a case study show an improvement in modeling accuracy of up to 8.74 dB in normalized mean square error by the adoption of the proposed configuration