In this paper we investigate an alternative characterization of the inverse AM/AM and AM/PM transfer curves of a high power amplifier (HPA), defining an adaptive algorithm for the optimum adjustment of the coefficients in a memoryless rational-based model used to pre-distort a base band signal with complex modulation in a quadrature non-linear model. Input signal statistics are unknown so its influence in the estimation is analyzed and compensation results are shown and compared with other previously proposed methods.