Generalization and Equilibrium in Generative Adversarial Nets (GANs)
Abstract
It is shown that training of generative adversarial network (GAN) may not have good generalization properties; e.g., training may appear successful but the trained distribution may be far from target distribution in standard metrics. However, generalization does occur for a weaker metric called neural net distance. It is also shown that an approximate pure equilibrium exists in the discriminator/generator game for a natural training objective (Wasserstein) when generator capacity and training set sizes are moderate. This existence of equilibrium inspires MIX+GAN protocol, which can be combined with any existing GAN training, and empirically shown to improve some of them.
Cite
Text
Arora et al. "Generalization and Equilibrium in Generative Adversarial Nets (GANs)." International Conference on Machine Learning, 2017.Markdown
[Arora et al. "Generalization and Equilibrium in Generative Adversarial Nets (GANs)." International Conference on Machine Learning, 2017.](https://mlanthology.org/icml/2017/arora2017icml-generalization/)BibTeX
@inproceedings{arora2017icml-generalization,
title = {{Generalization and Equilibrium in Generative Adversarial Nets (GANs)}},
author = {Arora, Sanjeev and Ge, Rong and Liang, Yingyu and Ma, Tengyu and Zhang, Yi},
booktitle = {International Conference on Machine Learning},
year = {2017},
pages = {224-232},
volume = {70},
url = {https://mlanthology.org/icml/2017/arora2017icml-generalization/}
}