Exploring Support Vector Regression for Modeling of GaN HEMT

This paper, for the first time, presents a support vector regression (SVR) based small signal modelling approach for a GaN High Electron Mobility Transistor (HEMT). It makes use of the nonlinear Gaussian or Radial Basis kernel function. The proposed technique has been demonstrated by modelling the small signal behavior of a 4×100μm GaN HEMT over a broad frequency range of 1-18 GHz with multi-biasing sets and measured S-parameters. An extremely good agreement has been found between the measured and the modelled performance for the complete frequency range.