Amortised Learning by Wake-Sleep

Li Wenliang, Theodore Moskovitz, Heishiro Kanagawa, Maneesh Sahani
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:10236-10247, 2020.

Abstract

Models that employ latent variables to capture structure in observed data lie at the heart of many current unsupervised learning algorithms, but exact maximum-likelihood learning for powerful and flexible latent-variable models is almost always intractable. Thus, state-of-the-art approaches either abandon the maximum-likelihood framework entirely, or else rely on a variety of variational approximations to the posterior distribution over the latents. Here, we propose an alternative approach that we call amortised learning. Rather than computing an approximation to the posterior over latents, we use a wake-sleep Monte-Carlo strategy to learn a function that directly estimates the maximum-likelihood parameter updates. Amortised learning is possible whenever samples of latents and observations can be simulated from the generative model, treating the model as a “black box”. We demonstrate its effectiveness on a wide range of complex models, including those with latents that are discrete or supported on non-Euclidean spaces.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-wenliang20a, title = {Amortised Learning by Wake-Sleep}, author = {Wenliang, Li and Moskovitz, Theodore and Kanagawa, Heishiro and Sahani, Maneesh}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {10236--10247}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/wenliang20a/wenliang20a.pdf}, url = {https://proceedings.mlr.press/v119/wenliang20a.html}, abstract = {Models that employ latent variables to capture structure in observed data lie at the heart of many current unsupervised learning algorithms, but exact maximum-likelihood learning for powerful and flexible latent-variable models is almost always intractable. Thus, state-of-the-art approaches either abandon the maximum-likelihood framework entirely, or else rely on a variety of variational approximations to the posterior distribution over the latents. Here, we propose an alternative approach that we call amortised learning. Rather than computing an approximation to the posterior over latents, we use a wake-sleep Monte-Carlo strategy to learn a function that directly estimates the maximum-likelihood parameter updates. Amortised learning is possible whenever samples of latents and observations can be simulated from the generative model, treating the model as a “black box”. We demonstrate its effectiveness on a wide range of complex models, including those with latents that are discrete or supported on non-Euclidean spaces.} }
Endnote
%0 Conference Paper %T Amortised Learning by Wake-Sleep %A Li Wenliang %A Theodore Moskovitz %A Heishiro Kanagawa %A Maneesh Sahani %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-wenliang20a %I PMLR %P 10236--10247 %U https://proceedings.mlr.press/v119/wenliang20a.html %V 119 %X Models that employ latent variables to capture structure in observed data lie at the heart of many current unsupervised learning algorithms, but exact maximum-likelihood learning for powerful and flexible latent-variable models is almost always intractable. Thus, state-of-the-art approaches either abandon the maximum-likelihood framework entirely, or else rely on a variety of variational approximations to the posterior distribution over the latents. Here, we propose an alternative approach that we call amortised learning. Rather than computing an approximation to the posterior over latents, we use a wake-sleep Monte-Carlo strategy to learn a function that directly estimates the maximum-likelihood parameter updates. Amortised learning is possible whenever samples of latents and observations can be simulated from the generative model, treating the model as a “black box”. We demonstrate its effectiveness on a wide range of complex models, including those with latents that are discrete or supported on non-Euclidean spaces.
APA
Wenliang, L., Moskovitz, T., Kanagawa, H. & Sahani, M.. (2020). Amortised Learning by Wake-Sleep. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:10236-10247 Available from https://proceedings.mlr.press/v119/wenliang20a.html.

Related Material