Asymmetric Variational Autoencoders

TL;DR

This paper introduces a novel framework for variational autoencoders that incorporates auxiliary variables, enabling more flexible inference models without the need for density evaluations of these variables, and demonstrates its effectiveness on density estimation tasks.

Contribution

It proposes a new approach to enrich variational families by adding auxiliary variables, allowing implicit densities and flexible mixture models in inference.

Findings

Effective in density estimation tasks

Enables implicit density construction with neural networks

Models rich probabilistic mixtures

Abstract

Variational inference for latent variable models is prevalent in various machine learning problems, typically solved by maximizing the Evidence Lower Bound (ELBO) of the true data likelihood with respect to a variational distribution. However, freely enriching the family of variational distribution is challenging since the ELBO requires variational likelihood evaluations of the latent variables. In this paper, we propose a novel framework to enrich the variational family by incorporating auxiliary variables to the variational family. The resulting inference network doesn't require density evaluations for the auxiliary variables and thus complex implicit densities over the auxiliary variables can be constructed by neural networks. It can be shown that the actual variational posterior of the proposed approach is essentially modeling a rich probabilistic mixture of simple variational posterior indexed by auxiliary variables, thus a flexible inference model can be built. Empirical evaluations on several density estimation tasks demonstrates the effectiveness of the proposed method.