373e4c5d8edfa8b74fd4b6791d0cf6dc-AuthorFeedback.pdf

Wewillreleaseamodularized1 implementation of MINs alongside all the tasks to make it easy to use and reproduce the results(R1). We will clarify this in the paper. The evaluation metric is simply the standard test classification accuracy for unseen test images.(ii)In We directly used implementations of GANs43 (details inApp. We used empirical densities and replace this choice byλ and τ (discussed in App.