Nonlinear Equivalent Circuit Model Base on BPNN for GaN HEMTs

Abstract

A Back-Propagation Neural Network (BPNN) nonlinear model of $GaN$ HEMT is proposed, which can adaptively fit the nonlinear parameter relationship, and reduce computation. The network weights are obtained automatically, and then the nonlinear mapping relation is determined. The comparison between BPNN model and test data proves the good consistency.