What is the advantage of GAN over simple bootstraping when the synthetic data is used as the training dataset for another ML/DL model? I\'m mostly interested in knowing the
