Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data

Yarin Gal, Yutian Chen, Zoubin Ghahramani
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:645-654, 2015.

Abstract

Multivariate categorical data occur in many applications of machine learning. One of the main difficulties with these vectors of categorical variables is sparsity. The number of possible observations grows exponentially with vector length, but dataset diversity might be poor in comparison. Recent models have gained significant improvement in supervised tasks with this data. These models embed observations in a continuous space to capture similarities between them. Building on these ideas we propose a Bayesian model for the unsupervised task of distribution estimation of multivariate categorical data. We model vectors of categorical variables as generated from a non-linear transformation of a continuous latent space. Non-linearity captures multi-modality in the distribution. The continuous representation addresses sparsity. Our model ties together many existing models, linking the linear categorical latent Gaussian model, the Gaussian process latent variable model, and Gaussian process classification. We derive inference for our model based on recent developments in sampling based variational inference. We show empirically that the model outperforms its linear and discrete counterparts in imputation tasks of sparse data.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-gala15, title = {Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data}, author = {Gal, Yarin and Chen, Yutian and Ghahramani, Zoubin}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {645--654}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/gala15.pdf}, url = {https://proceedings.mlr.press/v37/gala15.html}, abstract = {Multivariate categorical data occur in many applications of machine learning. One of the main difficulties with these vectors of categorical variables is sparsity. The number of possible observations grows exponentially with vector length, but dataset diversity might be poor in comparison. Recent models have gained significant improvement in supervised tasks with this data. These models embed observations in a continuous space to capture similarities between them. Building on these ideas we propose a Bayesian model for the unsupervised task of distribution estimation of multivariate categorical data. We model vectors of categorical variables as generated from a non-linear transformation of a continuous latent space. Non-linearity captures multi-modality in the distribution. The continuous representation addresses sparsity. Our model ties together many existing models, linking the linear categorical latent Gaussian model, the Gaussian process latent variable model, and Gaussian process classification. We derive inference for our model based on recent developments in sampling based variational inference. We show empirically that the model outperforms its linear and discrete counterparts in imputation tasks of sparse data.} }
Endnote
%0 Conference Paper %T Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data %A Yarin Gal %A Yutian Chen %A Zoubin Ghahramani %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-gala15 %I PMLR %P 645--654 %U https://proceedings.mlr.press/v37/gala15.html %V 37 %X Multivariate categorical data occur in many applications of machine learning. One of the main difficulties with these vectors of categorical variables is sparsity. The number of possible observations grows exponentially with vector length, but dataset diversity might be poor in comparison. Recent models have gained significant improvement in supervised tasks with this data. These models embed observations in a continuous space to capture similarities between them. Building on these ideas we propose a Bayesian model for the unsupervised task of distribution estimation of multivariate categorical data. We model vectors of categorical variables as generated from a non-linear transformation of a continuous latent space. Non-linearity captures multi-modality in the distribution. The continuous representation addresses sparsity. Our model ties together many existing models, linking the linear categorical latent Gaussian model, the Gaussian process latent variable model, and Gaussian process classification. We derive inference for our model based on recent developments in sampling based variational inference. We show empirically that the model outperforms its linear and discrete counterparts in imputation tasks of sparse data.
RIS
TY - CPAPER TI - Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data AU - Yarin Gal AU - Yutian Chen AU - Zoubin Ghahramani BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-gala15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 645 EP - 654 L1 - http://proceedings.mlr.press/v37/gala15.pdf UR - https://proceedings.mlr.press/v37/gala15.html AB - Multivariate categorical data occur in many applications of machine learning. One of the main difficulties with these vectors of categorical variables is sparsity. The number of possible observations grows exponentially with vector length, but dataset diversity might be poor in comparison. Recent models have gained significant improvement in supervised tasks with this data. These models embed observations in a continuous space to capture similarities between them. Building on these ideas we propose a Bayesian model for the unsupervised task of distribution estimation of multivariate categorical data. We model vectors of categorical variables as generated from a non-linear transformation of a continuous latent space. Non-linearity captures multi-modality in the distribution. The continuous representation addresses sparsity. Our model ties together many existing models, linking the linear categorical latent Gaussian model, the Gaussian process latent variable model, and Gaussian process classification. We derive inference for our model based on recent developments in sampling based variational inference. We show empirically that the model outperforms its linear and discrete counterparts in imputation tasks of sparse data. ER -
APA
Gal, Y., Chen, Y. & Ghahramani, Z.. (2015). Latent Gaussian Processes for Distribution Estimation of Multivariate Categorical Data. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:645-654 Available from https://proceedings.mlr.press/v37/gala15.html.

Related Material