InfoVAEGAN : Learning Joint Interpretable Representations by Information Maximization and Maximum Likelihood

Fei Ye; Adrian Gheorghe Bors

InfoVAEGAN : Learning Joint Interpretable Representations by Information Maximization and Maximum Likelihood

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Learning disentangled and interpretable representations is an
important step towards accomplishing comprehensive data
representations on the manifold. In this paper, we propose
a novel representation learning algorithm which combines
the inference abilities of Variational Autoencoders (VAE)
with the generalization capability of Generative Adversarial
Networks (GAN). The proposed model, called InfoVAEGAN,
consists of three networks : Encoder, Generator and
Discriminator. InfoVAEGAN aims to jointly learn discrete
and continuous interpretable representations in an unsupervised
manner by using two different data-free log-likelihood
functions onto the variables sampled from the generator’s
distribution. We propose a two-stage algorithm for optimizing
the inference network separately from the generator
training. Moreover, we enforce the learning of interpretable
representations through the maximization of the mutual information
between the existing latent variables and those created
through generative and inference processes.

Original language	English
Title of host publication	Proc. IEEE International Conference on Image Processing
Publisher	IEEE
Number of pages	5
Publication status	Published - 20 Sept 2021

Access to Document

ICIP2021_InfoVAEGANAccepted author manuscript, 1 MB

Cite this

@inproceedings{90d1d584e0de40708efe7d09d1ef150a,

title = "InfoVAEGAN : Learning Joint Interpretable Representations by Information Maximization and Maximum Likelihood",

abstract = "Learning disentangled and interpretable representations is animportant step towards accomplishing comprehensive datarepresentations on the manifold. In this paper, we proposea novel representation learning algorithm which combinesthe inference abilities of Variational Autoencoders (VAE)with the generalization capability of Generative AdversarialNetworks (GAN). The proposed model, called InfoVAEGAN,consists of three networks : Encoder, Generator andDiscriminator. InfoVAEGAN aims to jointly learn discreteand continuous interpretable representations in an unsupervisedmanner by using two different data-free log-likelihoodfunctions onto the variables sampled from the generator{\textquoteright}sdistribution. We propose a two-stage algorithm for optimizingthe inference network separately from the generatortraining. Moreover, we enforce the learning of interpretablerepresentations through the maximization of the mutual informationbetween the existing latent variables and those createdthrough generative and inference processes.",

author = "Fei Ye and Bors, {Adrian Gheorghe}",

year = "2021",

month = sep,

day = "20",

language = "English",

booktitle = "Proc. IEEE International Conference on Image Processing",

publisher = "IEEE",

address = "United States",

}

TY - GEN

T1 - InfoVAEGAN : Learning Joint Interpretable Representations by Information Maximization and Maximum Likelihood

AU - Ye, Fei

AU - Bors, Adrian Gheorghe

PY - 2021/9/20

Y1 - 2021/9/20

N2 - Learning disentangled and interpretable representations is animportant step towards accomplishing comprehensive datarepresentations on the manifold. In this paper, we proposea novel representation learning algorithm which combinesthe inference abilities of Variational Autoencoders (VAE)with the generalization capability of Generative AdversarialNetworks (GAN). The proposed model, called InfoVAEGAN,consists of three networks : Encoder, Generator andDiscriminator. InfoVAEGAN aims to jointly learn discreteand continuous interpretable representations in an unsupervisedmanner by using two different data-free log-likelihoodfunctions onto the variables sampled from the generator’sdistribution. We propose a two-stage algorithm for optimizingthe inference network separately from the generatortraining. Moreover, we enforce the learning of interpretablerepresentations through the maximization of the mutual informationbetween the existing latent variables and those createdthrough generative and inference processes.

AB - Learning disentangled and interpretable representations is animportant step towards accomplishing comprehensive datarepresentations on the manifold. In this paper, we proposea novel representation learning algorithm which combinesthe inference abilities of Variational Autoencoders (VAE)with the generalization capability of Generative AdversarialNetworks (GAN). The proposed model, called InfoVAEGAN,consists of three networks : Encoder, Generator andDiscriminator. InfoVAEGAN aims to jointly learn discreteand continuous interpretable representations in an unsupervisedmanner by using two different data-free log-likelihoodfunctions onto the variables sampled from the generator’sdistribution. We propose a two-stage algorithm for optimizingthe inference network separately from the generatortraining. Moreover, we enforce the learning of interpretablerepresentations through the maximization of the mutual informationbetween the existing latent variables and those createdthrough generative and inference processes.

M3 - Conference contribution

BT - Proc. IEEE International Conference on Image Processing

PB - IEEE

ER -