ICML 2017poster867 citations
Generalization and Equilibrium in Generative Adversarial Nets (GANs)
Sanjeev Arora, Rong Ge, Yingyu Liang, Tengyu Ma, Yi Zhang
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.
BibTeX
@InProceedings{pmlr-v70-arora17a,
title = {Generalization and Equilibrium in Generative Adversarial Nets ({GAN}s)},
author = {Sanjeev Arora and Rong Ge and Yingyu Liang and Tengyu Ma and Yi Zhang},
booktitle = {Proceedings of the 34th International Conference on Machine Learning},
pages = {224--232},
year = {2017},
editor = {Precup, Doina and Teh, Yee Whye},
volume = {70},
series = {Proceedings of Machine Learning Research},
month = {06--11 Aug},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v70/arora17a/arora17a.pdf},
url = {https://proceedings.mlr.press/v70/arora17a.html},
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.}
}