Question: Do you need two separate discriminators?

Open pfeatherstone opened this issue 5 years ago • 2 comments

I get similar-ish results when using the same discriminator for both cVAE-GAN and cLR-GAN. Is this a happy coincidence and a strike of luck? Is there some theoretical backing for having 2 discriminators?

Dec 01 '20 18:12 pfeatherstone

Just noticed the option use_same_D. Did that produce good results for you?

Dec 02 '20 07:12 pfeatherstone

Really wondering this too, the paper claims results were slightly better with two separate discriminators. However, because of #93 I am actually still unsure if the implementation for use_same_D is correct. If I understand #93 correctly, the discriminator when using use_same_D only receives half the updates, which makes me think this is where the findings that two separate discriminators are better came from. Anyone who could please provide some clarification on both this issue and #93? Thanks a lot!

Dec 16 '20 18:12 SanderGielisse