Vector Quantized Wasserstein Auto-Encoder

TL;DR

This paper introduces a generative approach to learning deep discrete representations using Wasserstein distance, improving upon VQ-VAE by enhancing codebook utilization and image generation quality.

Contribution

It proposes a novel Wasserstein Auto-Encoder framework for discrete representations, connecting it with clustering theory and demonstrating superior performance over existing VQ-VAE methods.

Findings

Achieves better codebook utilization

Improves image reconstruction quality

Outperforms other VQ-VAE variants

Abstract

Learning deep discrete latent presentations offers a promise of better symbolic and summarized abstractions that are more useful to subsequent downstream tasks. Inspired by the seminal Vector Quantized Variational Auto-Encoder (VQ-VAE), most of work in learning deep discrete representations has mainly focused on improving the original VQ-VAE form and none of them has studied learning deep discrete representations from the generative viewpoint. In this work, we study learning deep discrete representations from the generative viewpoint. Specifically, we endow discrete distributions over sequences of codewords and learn a deterministic decoder that transports the distribution over the sequences of codewords to the data distribution via minimizing a WS distance between them. We develop further theories to connect it with the clustering viewpoint of WS distance, allowing us to have a better and more controllable clustering solution. Finally, we empirically evaluate our method on several well-known benchmarks, where it achieves better qualitative and quantitative performances than the other VQ-VAE variants in terms of the codebook utilization and image reconstruction/generation.