Symmetric Wasserstein Autoencoders

TL;DR

This paper introduces Symmetric Wasserstein Autoencoders (SWAEs), a novel generative model leveraging optimal transport to symmetrically match data and latent distributions, improving data structure preservation and generative quality.

Contribution

The paper proposes a new autoencoder framework with a learnable prior that symmetrically matches data and latent distributions, enhancing structure preservation and generative performance.

Findings

SWAEs outperform state-of-the-art autoencoders in classification, reconstruction, and generation.

The symmetric matching approach preserves local data structure in the latent space.

Incorporating reconstruction loss improves both generation and reconstruction quality.

Abstract

Leveraging the framework of Optimal Transport, we introduce a new family of generative autoencoders with a learnable prior, called Symmetric Wasserstein Autoencoders (SWAEs). We propose to symmetrically match the joint distributions of the observed data and the latent representation induced by the encoder and the decoder. The resulting algorithm jointly optimizes the modelling losses in both the data and the latent spaces with the loss in the data space leading to the denoising effect. With the symmetric treatment of the data and the latent representation, the algorithm implicitly preserves the local structure of the data in the latent space. To further improve the quality of the latent representation, we incorporate a reconstruction loss into the objective, which significantly benefits both the generation and reconstruction. We empirically show the superior performance of SWAEs over the state-of-the-art generative autoencoders in terms of classification, reconstruction, and generation.