Skip to content

Chavdarova/LAGAN-Lookahead_Minimax

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LAGAN: Lookahead-Minmax

Source code of our paper Taming GANs with Lookahead–Minmax, ICLR 2021. Equal contribution with Matteo Pagliardini, and joint work with Sebastian Stich, François Fleuret, and Martin Jaggi.

To tackle the known challenges of minmax optimization of: (i) rotational joint vector field and (ii) sensitivity to noise induced by the stochastic gradient, we propose the Lookahead-Minmax algorithm. It consists of periodically taking an iterate that lies on a convex combination of the current and a past iterate.

lookahead-minmax illustration


The subdirectories contain:

  • lagan: the experiements on CIFAR10, SVHN and ImageNet.
  • mnist: the experiement on MNIST.
  • bilinear_game: the Colab-notebooks to reproduce the results on batch and stochastic bilinear game.

See also: paper, poster, and slides.

Citation

@inproceedings{chavdarova2021taming,
title={{Taming GANs with Lookahead-Minmax}},
author={Tatjana Chavdarova and Matteo Pagliardini and Sebastian U Stich and Fran{\c{c}}ois Fleuret and Martin Jaggi},
booktitle={International Conference on Learning Representations},
year={2021}
}

About

Source code for "Taming GANs with Lookahead–Minmax", ICLR 2021.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published