NeurIPS 2020

Improving GAN Training with Probability Ratio Clipping and Sample Reweighting

Meta Review

This is an interesting case. The R3 with the negative comments did not pointed out any specific flaws of the paper, so his review can not be taken into account. The whole idea is found interesting and simple, and this formulation has not been tested before in the GAN literature. The numerical experiments are quite broad and promising. The paper is well written, so I would argue that it can be accepted.