Theoretical Insights into the Use of Structural Similarity Index In
Generative Models and Inferential Autoencoders
B. Ghojogh, F. Karray, and M. Crowley. (2020)cite arxiv:2004.01864Comment: Accepted (to appear) in International Conference on Image Analysis and Recognition (ICIAR) 2020, Springer.
Abstract
Generative models and inferential autoencoders mostly make use of $\ell_2$
norm in their optimization objectives. In order to generate perceptually better
images, this short paper theoretically discusses how to use Structural
Similarity Index (SSIM) in generative models and inferential autoencoders. We
first review SSIM, SSIM distance metrics, and SSIM kernel. We show that the
SSIM kernel is a universal kernel and thus can be used in unconditional and
conditional generated moment matching networks. Then, we explain how to use
SSIM distance in variational and adversarial autoencoders and unconditional and
conditional Generative Adversarial Networks (GANs). Finally, we propose to use
SSIM distance rather than $\ell_2$ norm in least squares GAN.
Description
[2004.01864] Theoretical Insights into the Use of Structural Similarity Index In Generative Models and Inferential Autoencoders
%0 Journal Article
%1 ghojogh2020theoretical
%A Ghojogh, Benyamin
%A Karray, Fakhri
%A Crowley, Mark
%D 2020
%K generative-models index structure
%T Theoretical Insights into the Use of Structural Similarity Index In
Generative Models and Inferential Autoencoders
%U http://arxiv.org/abs/2004.01864
%X Generative models and inferential autoencoders mostly make use of $\ell_2$
norm in their optimization objectives. In order to generate perceptually better
images, this short paper theoretically discusses how to use Structural
Similarity Index (SSIM) in generative models and inferential autoencoders. We
first review SSIM, SSIM distance metrics, and SSIM kernel. We show that the
SSIM kernel is a universal kernel and thus can be used in unconditional and
conditional generated moment matching networks. Then, we explain how to use
SSIM distance in variational and adversarial autoencoders and unconditional and
conditional Generative Adversarial Networks (GANs). Finally, we propose to use
SSIM distance rather than $\ell_2$ norm in least squares GAN.
@article{ghojogh2020theoretical,
abstract = {Generative models and inferential autoencoders mostly make use of $\ell_2$
norm in their optimization objectives. In order to generate perceptually better
images, this short paper theoretically discusses how to use Structural
Similarity Index (SSIM) in generative models and inferential autoencoders. We
first review SSIM, SSIM distance metrics, and SSIM kernel. We show that the
SSIM kernel is a universal kernel and thus can be used in unconditional and
conditional generated moment matching networks. Then, we explain how to use
SSIM distance in variational and adversarial autoencoders and unconditional and
conditional Generative Adversarial Networks (GANs). Finally, we propose to use
SSIM distance rather than $\ell_2$ norm in least squares GAN.},
added-at = {2020-04-07T12:43:06.000+0200},
author = {Ghojogh, Benyamin and Karray, Fakhri and Crowley, Mark},
biburl = {https://www.bibsonomy.org/bibtex/24aec5acd30291bd3262aafbda4fc84bc/kirk86},
description = {[2004.01864] Theoretical Insights into the Use of Structural Similarity Index In Generative Models and Inferential Autoencoders},
interhash = {6f90f99dc7f29958f48210440561f709},
intrahash = {4aec5acd30291bd3262aafbda4fc84bc},
keywords = {generative-models index structure},
note = {cite arxiv:2004.01864Comment: Accepted (to appear) in International Conference on Image Analysis and Recognition (ICIAR) 2020, Springer},
timestamp = {2020-04-07T12:43:06.000+0200},
title = {Theoretical Insights into the Use of Structural Similarity Index In
Generative Models and Inferential Autoencoders},
url = {http://arxiv.org/abs/2004.01864},
year = 2020
}