3 ms·
Anyone know why a GAN was used here? The discriminator makes sense, but what is the purpose of the generator for the given experiment? Why not cut out the gener
by notJim 3y ago
Anyone know why a GAN was used here? The discriminator makes sense, but what is the purpose of the generator for the given experiment? Why not cut out the generator, since the discriminator is trained separately on speech data anyway.
- space_fountain 3y agoUsually 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