$\rho$-VAE: Autoregressive parametrization of the VAE encoder

TL;DR

This paper introduces $ ho$-VAE, a simple yet effective autoregressive modification to the VAE encoder that improves image generation quality by better modeling correlations in the data, with minimal additional parameters.

Contribution

It proposes a drop-in autoregressive Gaussian posterior for VAEs, enhancing correlation modeling while reducing parameter count and maintaining closed-form KL divergence.

Findings

Improves image generation quality across various setups.

Requires only two additional scalar parameters.

No extra parameter tuning needed.

Abstract

We make a minimal, but very effective alteration to the VAE model. This is about a drop-in replacement for the (sample-dependent) approximate posterior to change it from the standard white Gaussian with diagonal covariance to the first-order autoregressive Gaussian. We argue that this is a more reasonable choice to adopt for natural signals like images, as it does not force the existing correlation in the data to disappear in the posterior. Moreover, it allows more freedom for the approximate posterior to match the true posterior. This allows for the repararametrization trick, as well as the KL-divergence term to still have closed-form expressions, obviating the need for its sample-based estimation. Although providing more freedom to adapt to correlated distributions, our parametrization has even less number of parameters than the diagonal covariance, as it requires only two scalars, $\rho$ and $s$, to characterize correlation and scaling, respectively. As validated by the experiments, our proposition noticeably and consistently improves the quality of image generation in a plug-and-play manner, needing no further parameter tuning, and across all setups. The code to reproduce our experiments is available at \url{https://github.com/sssohrab/rho_VAE/}.