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.

Sign up for GitHub

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

Jump to bottom

Sigmoid for lsgan #70

Closed
fanpengustc opened this issue Feb 10, 2018 · 4 comments
Closed

Sigmoid for lsgan #70

fanpengustc opened this issue Feb 10, 2018 · 4 comments

Comments

@fanpengustc
Copy link

Why sigmoid is not used in Discriminator for lsgan, if that has some special meaning？
@junyanz
Copy link
Owner

junyanz commented Feb 10, 2018

Sigmoid activation is removed in the LSGAN. Otherwise, there might be vanishing gradient issue. See this post for more details.

@fanpengustc
Copy link
Author

Thanks for your answer! But in the post, I did not find the sigmoid activation is removed!

@junyanz
Copy link
Owner

junyanz commented Feb 15, 2018

You can find details in the 'LSGAN implementation in Pytorch' section. There is a comment saying # No sigmoid.

@fanpengustc
Copy link
Author

Thanks!

@edchengg edchengg mentioned this issue Jun 18, 2018
Closed
@neel-dey neel-dey mentioned this issue Oct 17, 2019
Closed
simontomaskarlsson pushed a commit to simontomaskarlsson/GAN-MRI that referenced this issue Oct 23, 2019
Update CycleGAN.py
f63e147 
Removing sigmoid in discriminator model to avoid near-fp32 machine epsilon discriminator cost, as in junyanz/CycleGAN#70
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@junyanz @fanpengustc