The algorithm provides a method for compression of GANs, which has not been addressed previously. It is simple and has good empirical performance. While the paper provides strong empirical performance for a relevant task, the main weakness of the paper is that it lacks in terms of theoretical motivation and theoretical insights.