https://github.com/locuslab/gradient_regularized_gan

Code for "Gradient descent GAN optimization is locally stable"
https://github.com/locuslab/gradient_regularized_gan

Last synced: about 1 year ago
JSON representation

Code for "Gradient descent GAN optimization is locally stable"

• Host: GitHub
• URL: https://github.com/locuslab/gradient_regularized_gan
• Owner: locuslab
• Created: 2017-11-03T01:15:18.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2017-11-05T09:37:03.000Z (over 8 years ago)
• Last Synced: 2025-04-02T20:11:24.952Z (about 1 year ago)
• Language: Python
• Homepage:
• Size: 5.86 KB
• Stars: 21
• Watchers: 6
• Forks: 8
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Gradient Regularized GAN

Code for the regularizer proposed in the NIPS 2017 paper on "Gradient descent GAN optimization is locally stable"

arXiv:1706.04156

The code for the Gaussian experiments is based on https://github.com/poolio/unrolled_gan but modified to reflect the parameters specified in the Unrolled GANs paper arxiv:1611.02163. The experiments for our paper were performed on Tensorflow 1.1.0 and Keras 1.2. 

Run ```python .py``` where `````` is either ```gaussian-toy-unrolled.py``` or ```gaussian-toy-regularized.py``` to run the unrolled GAN and the gradient-norm-regularized GAN respectively.