Dear Tolstikhin
Thanks for your solid work, but I want to consult you about the MMD loss function during my specific training: I found that MMD loss function is negative when I was training other datasets except for MNIST and CelebA. Why did that circumstance happen and what did that mean? How should I tackle the negative loss function?
Dear Tolstikhin Thanks for your solid work, but I want to consult you about the MMD loss function during my specific training: I found that MMD loss function is negative when I was training other datasets except for MNIST and CelebA. Why did that circumstance happen and what did that mean? How should I tackle the negative loss function?