Self-Supervised Variational Auto-Encoders

TL;DR

This paper introduces self-supervised Variational Auto-Encoders (selfVAE), a new generative model framework that simplifies training and enhances data reconstruction, with applications in data compression and generation.

Contribution

The paper proposes a novel self-supervised VAE framework utilizing deterministic and discrete posteriors, including hierarchical architectures, improving flexibility and performance over traditional VAEs.

Findings

Effective data reconstruction with flexible trade-offs between memory and quality.

Hierarchical selfVAE architecture outperforms standard VAEs on benchmark datasets.

Demonstrated applicability in image data compression and generation tasks.

Abstract

Density estimation, compression and data generation are crucial tasks in artificial intelligence. Variational Auto-Encoders (VAEs) constitute a single framework to achieve these goals. Here, we present a novel class of generative models, called self-supervised Variational Auto-Encoder (selfVAE), that utilizes deterministic and discrete variational posteriors. This class of models allows to perform both conditional and unconditional sampling, while simplifying the objective function. First, we use a single self-supervised transformation as a latent variable, where a transformation is either downscaling or edge detection. Next, we consider a hierarchical architecture, i.e., multiple transformations, and we show its benefits compared to the VAE. The flexibility of selfVAE in data reconstruction finds a particularly interesting use case in data compression tasks, where we can trade-off memory for better data quality, and vice-versa. We present performance of our approach on three benchmark image data (Cifar10, Imagenette64, and CelebA).