Inference Suboptimality in Variational Autoencoders

TL;DR

This paper investigates why variational autoencoders often produce suboptimal inference, highlighting the role of recognition network quality and how generator learning influences approximation expressiveness.

Contribution

It reveals that recognition network limitations, not distribution complexity, primarily cause inference suboptimality, and explores how generator training affects approximation expressiveness.

Findings

Recognition network imperfections cause divergence from true posterior.

Generator learning adapts to the recognition network, influencing inference quality.

Expressiveness parameters aid in generalizing inference beyond complexity improvements.

Abstract

Amortized inference allows latent-variable models trained via variational learning to scale to large datasets. The quality of approximate inference is determined by two factors: a) the capacity of the variational distribution to match the true posterior and b) the ability of the recognition network to produce good variational parameters for each datapoint. We examine approximate inference in variational autoencoders in terms of these factors. We find that divergence from the true posterior is often due to imperfect recognition networks, rather than the limited complexity of the approximating distribution. We show that this is due partly to the generator learning to accommodate the choice of approximation. Furthermore, we show that the parameters used to increase the expressiveness of the approximation play a role in generalizing inference rather than simply improving the complexity of the approximation.