Usually discriminators aren’t trained separately. Training them together makes the initial task for the generative side easier since it doesn’t have to be near

Usually discriminators aren’t trained separately. Training them together makes the initial task for the generative side easier since it doesn’t have to be near perfect right of the bay. It can learn the obvious mistakes at the same time the discriminator is learning to detect them