representation learning and we share the reviewers ' enthusiasm to explore further implications of our method, the first to

We thank the reviewers for their time and comments. We show similar results on the 3d Chairs dataset, with style as the label and yaw angle as the desired invariance.