Amortised Learning by Wake-Sleep

TL;DR

This paper introduces amortised learning, a novel approach for training latent-variable models by directly estimating parameter updates through a wake-sleep Monte Carlo method, bypassing complex posterior approximations.

Contribution

It proposes a new amortised learning method that uses wake-sleep Monte Carlo strategies to estimate maximum-likelihood updates without explicit posterior inference.

Findings

Effective on models with discrete latents

Works with non-Euclidean latent spaces

Outperforms traditional variational methods

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.