Auto-Encoding Total Correlation Explanation

Shuyang Gao, Rob Brekelmans, Greg Ver Steeg, Aram Galstyan
Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, PMLR 89:1157-1166, 2019.

Abstract

Advances in unsupervised learning enable reconstruction and generation of samples from complex distributions, but this success is marred by the inscrutability of the representations learned. We propose an information-theoretic approach to characterizing disentanglement and dependence in representation learning using multivariate mutual information, also called total correlation. The principle of Total Cor-relation Ex-planation (CorEx) has motivated successful unsupervised learning applications across a variety of domains but under some restrictive assumptions. Here we relax those restrictions by introducing a flexible variational lower bound to CorEx. Surprisingly, we find this lower bound is equivalent to the one in variational autoencoders (VAE) under certain conditions. This information-theoretic view of VAE deepens our understanding of hierarchical VAE and motivates a new algorithm, AnchorVAE, that makes latent codes more interpretable through information maximization and enables generation of richer and more realistic samples.

Cite this Paper


BibTeX
@InProceedings{pmlr-v89-gao19a, title = {Auto-Encoding Total Correlation Explanation}, author = {Gao, Shuyang and Brekelmans, Rob and Steeg, Greg Ver and Galstyan, Aram}, booktitle = {Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics}, pages = {1157--1166}, year = {2019}, editor = {Chaudhuri, Kamalika and Sugiyama, Masashi}, volume = {89}, series = {Proceedings of Machine Learning Research}, month = {16--18 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v89/gao19a/gao19a.pdf}, url = {https://proceedings.mlr.press/v89/gao19a.html}, abstract = {Advances in unsupervised learning enable reconstruction and generation of samples from complex distributions, but this success is marred by the inscrutability of the representations learned. We propose an information-theoretic approach to characterizing disentanglement and dependence in representation learning using multivariate mutual information, also called total correlation. The principle of Total Cor-relation Ex-planation (CorEx) has motivated successful unsupervised learning applications across a variety of domains but under some restrictive assumptions. Here we relax those restrictions by introducing a flexible variational lower bound to CorEx. Surprisingly, we find this lower bound is equivalent to the one in variational autoencoders (VAE) under certain conditions. This information-theoretic view of VAE deepens our understanding of hierarchical VAE and motivates a new algorithm, AnchorVAE, that makes latent codes more interpretable through information maximization and enables generation of richer and more realistic samples.} }
Endnote
%0 Conference Paper %T Auto-Encoding Total Correlation Explanation %A Shuyang Gao %A Rob Brekelmans %A Greg Ver Steeg %A Aram Galstyan %B Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Masashi Sugiyama %F pmlr-v89-gao19a %I PMLR %P 1157--1166 %U https://proceedings.mlr.press/v89/gao19a.html %V 89 %X Advances in unsupervised learning enable reconstruction and generation of samples from complex distributions, but this success is marred by the inscrutability of the representations learned. We propose an information-theoretic approach to characterizing disentanglement and dependence in representation learning using multivariate mutual information, also called total correlation. The principle of Total Cor-relation Ex-planation (CorEx) has motivated successful unsupervised learning applications across a variety of domains but under some restrictive assumptions. Here we relax those restrictions by introducing a flexible variational lower bound to CorEx. Surprisingly, we find this lower bound is equivalent to the one in variational autoencoders (VAE) under certain conditions. This information-theoretic view of VAE deepens our understanding of hierarchical VAE and motivates a new algorithm, AnchorVAE, that makes latent codes more interpretable through information maximization and enables generation of richer and more realistic samples.
APA
Gao, S., Brekelmans, R., Steeg, G.V. & Galstyan, A.. (2019). Auto-Encoding Total Correlation Explanation. Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 89:1157-1166 Available from https://proceedings.mlr.press/v89/gao19a.html.

Related Material