English

Variational Information Bottleneck on Vector Quantized Autoencoders

Machine Learning 2018-08-06 v1 Machine Learning

Abstract

In this paper, we provide an information-theoretic interpretation of the Vector Quantized-Variational Autoencoder (VQ-VAE). We show that the loss function of the original VQ-VAE can be derived from the variational deterministic information bottleneck (VDIB) principle. On the other hand, the VQ-VAE trained by the Expectation Maximization (EM) algorithm can be viewed as an approximation to the variational information bottleneck(VIB) principle.

Keywords

Cite

@article{arxiv.1808.01048,
  title  = {Variational Information Bottleneck on Vector Quantized Autoencoders},
  author = {Hanwei Wu and Markus Flierl},
  journal= {arXiv preprint arXiv:1808.01048},
  year   = {2018}
}
R2 v1 2026-06-23T03:23:24.474Z