Reviews: One-Shot Unsupervised Cross Domain Translation

In overall, I think this paper proposes a well-designed two steps learning pipeline for one-shot unsupervised image translation. But the explanations about selective backpropagation in the rebuttal are still not clear to me. According to Eq.8-14, it seems that G S and E S are not updated in phase II. But according to the Tab. 1 and the rebuttal, they seem to be selectively updated. I strongly suggest the authors to explain the details and motivation in the method part if this paper is accepted.