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

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/}.

Submitted to NeurIPS workshop on Bayesian deep learning

References in corpus (9)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Semi-Supervised Learning with Deep Generative Models
• Estimating divergence functionals and the likelihood ratio by convex risk minimization
• Recent Advances in Autoencoder-Based Representation Learning
• InfoVAE: Information Maximizing Variational Autoencoders
• Variational Lossy Autoencoder
• Diagnosing and Enhancing VAE Models
• VAE with a VampPrior