Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data

Luigi Antelmi, Nicholas Ayache, Philippe Robert, Marco Lorenzi
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:302-311, 2019.

Abstract

Interpretable modeling of heterogeneous data channels is essential in medical applications, for example when jointly analyzing clinical scores and medical images. Variational Autoencoders (VAE) are powerful generative models that learn representations of complex data. The flexibility of VAE may come at the expense of lack of interpretability in describing the joint relationship between heterogeneous data. To tackle this problem, in this work we extend the variational framework of VAE to bring parsimony and interpretability when jointly account for latent relationships across multiple channels. In the latent space, this is achieved by constraining the variational distribution of each channel to a common target prior. Parsimonious latent representations are enforced by variational dropout. Experiments on synthetic data show that our model correctly identifies the prescribed latent dimensions and data relationships across multiple testing scenarios. When applied to imaging and clinical data, our method allows to identify the joint effect of age and pathology in describing clinical condition in a large scale clinical cohort.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-antelmi19a, title = {Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data}, author = {Antelmi, Luigi and Ayache, Nicholas and Robert, Philippe and Lorenzi, Marco}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {302--311}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/antelmi19a/antelmi19a.pdf}, url = {https://proceedings.mlr.press/v97/antelmi19a.html}, abstract = {Interpretable modeling of heterogeneous data channels is essential in medical applications, for example when jointly analyzing clinical scores and medical images. Variational Autoencoders (VAE) are powerful generative models that learn representations of complex data. The flexibility of VAE may come at the expense of lack of interpretability in describing the joint relationship between heterogeneous data. To tackle this problem, in this work we extend the variational framework of VAE to bring parsimony and interpretability when jointly account for latent relationships across multiple channels. In the latent space, this is achieved by constraining the variational distribution of each channel to a common target prior. Parsimonious latent representations are enforced by variational dropout. Experiments on synthetic data show that our model correctly identifies the prescribed latent dimensions and data relationships across multiple testing scenarios. When applied to imaging and clinical data, our method allows to identify the joint effect of age and pathology in describing clinical condition in a large scale clinical cohort.} }
Endnote
%0 Conference Paper %T Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data %A Luigi Antelmi %A Nicholas Ayache %A Philippe Robert %A Marco Lorenzi %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-antelmi19a %I PMLR %P 302--311 %U https://proceedings.mlr.press/v97/antelmi19a.html %V 97 %X Interpretable modeling of heterogeneous data channels is essential in medical applications, for example when jointly analyzing clinical scores and medical images. Variational Autoencoders (VAE) are powerful generative models that learn representations of complex data. The flexibility of VAE may come at the expense of lack of interpretability in describing the joint relationship between heterogeneous data. To tackle this problem, in this work we extend the variational framework of VAE to bring parsimony and interpretability when jointly account for latent relationships across multiple channels. In the latent space, this is achieved by constraining the variational distribution of each channel to a common target prior. Parsimonious latent representations are enforced by variational dropout. Experiments on synthetic data show that our model correctly identifies the prescribed latent dimensions and data relationships across multiple testing scenarios. When applied to imaging and clinical data, our method allows to identify the joint effect of age and pathology in describing clinical condition in a large scale clinical cohort.
APA
Antelmi, L., Ayache, N., Robert, P. & Lorenzi, M.. (2019). Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:302-311 Available from https://proceedings.mlr.press/v97/antelmi19a.html.

Related Material