A new digital predistortion (DPD) technique based on direct learning architecture is proposed for linearizing power amplifiers (PAs) with nonlinear memory effects. It is based on the non-uniform memory polynomial (MP) model and with quadratically interpolated LUT. Compared with the conventional MP model, the non-uniform MP model achieves the same modeling accuracy with less coefficients. Additionally, quadratically interpolated LUT is adopted to realize a part of mathematical computations in the predistortion. The proposed solution is validated on a real PA ZFL-2500. It reduces the number of model’s coefficients and achieves a spectrum regrowth suppression of about 15dB with table size of only 8.