A comparative study of learning architecture for digital predistortion

In the paper, we conduct a comparative study of three learning architectures with simulations and experiments for adaptive digital pre-distortion: direct learning (DLA), indirect direct learning (ILA), and intuitive direct learning (iDLA). Our study shows the iDLA achieves the same linearization performance as the DLA, and has better performance than the ILA (1 ~ 2 dB for our tests), which suffers more from feedback noise. The iDLA has the same complexity as the ILA, and only half of the DLA complexity.