The Variational Homoencoder: Learning to learn high capacity generative models from few examples

Luke B. Hewitt, Maxwell I. Nye, Andreea Gane, Tommi Jaakkola, Joshua B. Tenenbaum
Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, PMLR R16:987-996, 2018.

Abstract

Hierarchical Bayesian methods can unify many related tasks (e.g. k-shot classification, conditional and unconditional generation) as inference within a single generative model. However, when this generative model is ex- pressed as a powerful neural network such as a PixelCNN, we show that existing learning techniques typically fail to effectively use la- tent variables. To address this, we develop a modification of the Variational Autoencoder in which encoded observations are decoded to new elements from the same class. This technique, which we call a Variational Ho- moencoder (VHE), produces a hierarchical la- tent variable model which better utilises la- tent variables. We use the VHE framework to learn a hierarchical PixelCNN on the Omniglot dataset, which outperforms all existing models on test set likelihood and achieves strong per- formance on one-shot generation and classifi- cation tasks. We additionally validate the VHE on natural images from the YouTube Faces database. Finally, we develop extensions of the model that apply to richer dataset structures such as factorial and hierarchical categories.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR16-hewitt18a, title = {The Variational Homoencoder: Learning to learn high capacity generative models from few examples}, author = {Hewitt, Luke B. and Nye, Maxwell I. and Gane, Andreea and Jaakkola, Tommi and Tenenbaum, Joshua B.}, booktitle = {Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence}, pages = {987--996}, year = {2018}, editor = {Globerson, Amir and Silva, Ricardo}, volume = {R16}, series = {Proceedings of Machine Learning Research}, month = {06--10 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r16/main/assets/hewitt18a/hewitt18a.pdf}, url = {https://proceedings.mlr.press/r16/hewitt18a.html}, abstract = {Hierarchical Bayesian methods can unify many related tasks (e.g. k-shot classification, conditional and unconditional generation) as inference within a single generative model. However, when this generative model is ex- pressed as a powerful neural network such as a PixelCNN, we show that existing learning techniques typically fail to effectively use la- tent variables. To address this, we develop a modification of the Variational Autoencoder in which encoded observations are decoded to new elements from the same class. This technique, which we call a Variational Ho- moencoder (VHE), produces a hierarchical la- tent variable model which better utilises la- tent variables. We use the VHE framework to learn a hierarchical PixelCNN on the Omniglot dataset, which outperforms all existing models on test set likelihood and achieves strong per- formance on one-shot generation and classifi- cation tasks. We additionally validate the VHE on natural images from the YouTube Faces database. Finally, we develop extensions of the model that apply to richer dataset structures such as factorial and hierarchical categories.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T The Variational Homoencoder: Learning to learn high capacity generative models from few examples %A Luke B. Hewitt %A Maxwell I. Nye %A Andreea Gane %A Tommi Jaakkola %A Joshua B. Tenenbaum %B Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2018 %E Amir Globerson %E Ricardo Silva %F pmlr-vR16-hewitt18a %I PMLR %P 987--996 %U https://proceedings.mlr.press/r16/hewitt18a.html %V R16 %X Hierarchical Bayesian methods can unify many related tasks (e.g. k-shot classification, conditional and unconditional generation) as inference within a single generative model. However, when this generative model is ex- pressed as a powerful neural network such as a PixelCNN, we show that existing learning techniques typically fail to effectively use la- tent variables. To address this, we develop a modification of the Variational Autoencoder in which encoded observations are decoded to new elements from the same class. This technique, which we call a Variational Ho- moencoder (VHE), produces a hierarchical la- tent variable model which better utilises la- tent variables. We use the VHE framework to learn a hierarchical PixelCNN on the Omniglot dataset, which outperforms all existing models on test set likelihood and achieves strong per- formance on one-shot generation and classifi- cation tasks. We additionally validate the VHE on natural images from the YouTube Faces database. Finally, we develop extensions of the model that apply to richer dataset structures such as factorial and hierarchical categories. %Z Reissued by PMLR on 04 October 2026.
APA
Hewitt, L.B., Nye, M.I., Gane, A., Jaakkola, T. & Tenenbaum, J.B.. (2018). The Variational Homoencoder: Learning to learn high capacity generative models from few examples. Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R16:987-996 Available from https://proceedings.mlr.press/r16/hewitt18a.html. Reissued by PMLR on 04 October 2026.

Related Material