Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Change the discriminator objective function for ACGAN #85

merged 1 commit into from Apr 1, 2019


None yet
2 participants
Copy link

commented Oct 10, 2018

From the paper in discussion:

The discriminator gives both a probability distribution over sources and a probability distribution over the class labels, P(S | X), P(C | X) = D(X)

The discriminator should (along with P(S | X) ) give the probability that the image in analysis belongs to a certain class regardless of whether it is fake or original.
(Also, the real-fake information is already given in the other output, i.e. P(S | X) ).

The current implementation might give better results during the initial iterations, as at the beginning the generator output is mainly noise which thus might hinder the learning of the discriminator.
However, in my opinion as the process continues the paper idea works better, as more useful information is given from the discriminator to the generator for its learning.

@eriklindernoren eriklindernoren merged commit 44d3320 into eriklindernoren:master Apr 1, 2019

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.