question-mark
Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Discriminator dominates over generator

See original GitHub issue

I could get audible results at epoch>350, but they don’t look good. Also, d_loss gets too low and g_loss gets too high.

Screen Shot 2019-10-19 at 7 39 38 PM

Perhaps this could be caused by:

Issue Analytics

  • State:closed
  • Created 4 years ago
  • Comments:25 (14 by maintainers)

github_iconTop GitHub Comments

1reaction
seungwonparkcommented, Oct 25, 2019

This is getting interesting! After about 2000 epochs of training, discriminator finds it almost impossible to discriminate real/fake.

image

1reaction
seungwonparkcommented, Oct 24, 2019

Getting good quality audio while generator loss is still going up, too. (Audio samples at http://swpark.me/melgan/)

@rishikksh20 Even if generator loss is going up, it doesn’t seem that discriminator is overly better than generator. How do you think? Shall we close this issue?

image

Read more comments on GitHub >

github_iconTop Results From Across the Web

Common Problems | Machine Learning - Google Developers
Each iteration of generator over-optimizes for a particular discriminator, and the discriminator never manages to learn its way out of the trap.
Read more >
How to Identify and Diagnose GAN Failure Modes
Discriminator accuracy on real and fake images is expected to sit around 80%. Variance of generator and discriminator loss is expected to remain ......
Read more >
GANs Failure Modes: How to Identify and Monitor Them
An aggressive modifier is bad news for training GANs. It results in the inability to find the equilibrium between generator loss and discriminator...
Read more >
Why use the output of the generator to train the discriminator ...
The main reason that the discriminator is trained concurrently with the generator is to provide (at least in theory) a smooth and gradual ......
Read more >
python - How to balance the generator and the discriminator ...
that is, the gradients on the Generator are higher and start to decrease after a while, and in the meanwhile the gradients on...
Read more >

github_iconTop Related Medium Post

No results found

github_iconTop Related StackOverflow Question

No results found

github_iconTroubleshoot Live Code

Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start Free

github_iconTop Related Reddit Thread

No results found

github_iconTop Related Hackernoon Post

No results found

github_iconTop Related Tweet

No results found

github_iconTop Related Dev.to Post

No results found

github_iconTop Related Hashnode Post

No results found