The goal of this paper is not to introduce a single algorithm or method, but
to make theoretical steps towards fully understanding the training dynamics of
generative adversarial networks. In order to substantiate our theoretical
analysis, we perform targeted experiments to verify our assumptions, illustrate
our claims, and quantify the phenomena. This paper is divided into three
sections. The first section introduces the problem at hand. The second section
is dedicated to studying and proving rigorously the problems including
instability and saturation that arize when training generative adversarial
networks. The third section examines a practical and theoretically grounded
direction towards solving these problems, while introducing new tools to study
them.
Описание
Towards Principled Methods for Training Generative Adversarial Networks
%0 Generic
%1 arjovsky2017towards
%A Arjovsky, Martin
%A Bottou, Léon
%D 2017
%K cs.LG stat.ML
%T Towards Principled Methods for Training Generative Adversarial Networks
%U http://arxiv.org/abs/1701.04862
%X The goal of this paper is not to introduce a single algorithm or method, but
to make theoretical steps towards fully understanding the training dynamics of
generative adversarial networks. In order to substantiate our theoretical
analysis, we perform targeted experiments to verify our assumptions, illustrate
our claims, and quantify the phenomena. This paper is divided into three
sections. The first section introduces the problem at hand. The second section
is dedicated to studying and proving rigorously the problems including
instability and saturation that arize when training generative adversarial
networks. The third section examines a practical and theoretically grounded
direction towards solving these problems, while introducing new tools to study
them.
@misc{arjovsky2017towards,
abstract = {The goal of this paper is not to introduce a single algorithm or method, but
to make theoretical steps towards fully understanding the training dynamics of
generative adversarial networks. In order to substantiate our theoretical
analysis, we perform targeted experiments to verify our assumptions, illustrate
our claims, and quantify the phenomena. This paper is divided into three
sections. The first section introduces the problem at hand. The second section
is dedicated to studying and proving rigorously the problems including
instability and saturation that arize when training generative adversarial
networks. The third section examines a practical and theoretically grounded
direction towards solving these problems, while introducing new tools to study
them.},
added-at = {2021-08-06T09:18:32.000+0200},
author = {Arjovsky, Martin and Bottou, Léon},
biburl = {https://www.bibsonomy.org/bibtex/2195eedc1486a24289902484164f2ed03/aerover},
description = {Towards Principled Methods for Training Generative Adversarial Networks},
interhash = {d5b86d070393fda09fb3e685f13a3811},
intrahash = {195eedc1486a24289902484164f2ed03},
keywords = {cs.LG stat.ML},
note = {cite arxiv:1701.04862},
timestamp = {2022-06-06T15:48:02.000+0200},
title = {Towards Principled Methods for Training Generative Adversarial Networks},
url = {http://arxiv.org/abs/1701.04862},
year = 2017
}