We propose a new framework for estimating generative models via an
adversarial process, in which we simultaneously train two models: a generative
model G that captures the data distribution, and a discriminative model D that
estimates the probability that a sample came from the training data rather than
G. The training procedure for G is to maximize the probability of D making a
mistake. This framework corresponds to a minimax two-player game. In the space
of arbitrary functions G and D, a unique solution exists, with G recovering the
training data distribution and D equal to 1/2 everywhere. In the case where G
and D are defined by multilayer perceptrons, the entire system can be trained
with backpropagation. There is no need for any Markov chains or unrolled
approximate inference networks during either training or generation of samples.
Experiments demonstrate the potential of the framework through qualitative and
quantitative evaluation of the generated samples.
%0 Generic
%1 goodfellow2014generative
%A Goodfellow, Ian J.
%A Pouget-Abadie, Jean
%A Mirza, Mehdi
%A Xu, Bing
%A Warde-Farley, David
%A Ozair, Sherjil
%A Courville, Aaron
%A Bengio, Yoshua
%D 2014
%K GAN
%T Generative Adversarial Networks
%U http://arxiv.org/abs/1406.2661
%X We propose a new framework for estimating generative models via an
adversarial process, in which we simultaneously train two models: a generative
model G that captures the data distribution, and a discriminative model D that
estimates the probability that a sample came from the training data rather than
G. The training procedure for G is to maximize the probability of D making a
mistake. This framework corresponds to a minimax two-player game. In the space
of arbitrary functions G and D, a unique solution exists, with G recovering the
training data distribution and D equal to 1/2 everywhere. In the case where G
and D are defined by multilayer perceptrons, the entire system can be trained
with backpropagation. There is no need for any Markov chains or unrolled
approximate inference networks during either training or generation of samples.
Experiments demonstrate the potential of the framework through qualitative and
quantitative evaluation of the generated samples.
@misc{goodfellow2014generative,
abstract = {We propose a new framework for estimating generative models via an
adversarial process, in which we simultaneously train two models: a generative
model G that captures the data distribution, and a discriminative model D that
estimates the probability that a sample came from the training data rather than
G. The training procedure for G is to maximize the probability of D making a
mistake. This framework corresponds to a minimax two-player game. In the space
of arbitrary functions G and D, a unique solution exists, with G recovering the
training data distribution and D equal to 1/2 everywhere. In the case where G
and D are defined by multilayer perceptrons, the entire system can be trained
with backpropagation. There is no need for any Markov chains or unrolled
approximate inference networks during either training or generation of samples.
Experiments demonstrate the potential of the framework through qualitative and
quantitative evaluation of the generated samples.},
added-at = {2017-08-18T10:33:29.000+0200},
author = {Goodfellow, Ian J. and Pouget-Abadie, Jean and Mirza, Mehdi and Xu, Bing and Warde-Farley, David and Ozair, Sherjil and Courville, Aaron and Bengio, Yoshua},
biburl = {https://www.bibsonomy.org/bibtex/2c356abc854ece0e874b56391b9eb8f0d/daschloer},
description = {Generative Adversarial Networks},
interhash = {2f4fdea569fc1ba2057a9af75bb95bc4},
intrahash = {c356abc854ece0e874b56391b9eb8f0d},
keywords = {GAN},
note = {cite arxiv:1406.2661},
timestamp = {2017-10-04T16:28:59.000+0200},
title = {Generative Adversarial Networks},
url = {http://arxiv.org/abs/1406.2661},
year = 2014
}