SayIt

So, this just means, this is an DP, I think, such that we have strong guarantees by differential privacy. So, we either train generators with differential privacy, so basically, the parameters cross the privacy boundary…