English

Variational Bayesian Quantization

Image and Video Processing 2020-09-09 v2 Computer Vision and Pattern Recognition Machine Learning Machine Learning

Abstract

We propose a novel algorithm for quantizing continuous latent representations in trained models. Our approach applies to deep probabilistic models, such as variational autoencoders (VAEs), and enables both data and model compression. Unlike current end-to-end neural compression methods that cater the model to a fixed quantization scheme, our algorithm separates model design and training from quantization. Consequently, our algorithm enables "plug-and-play" compression with variable rate-distortion trade-off, using a single trained model. Our algorithm can be seen as a novel extension of arithmetic coding to the continuous domain, and uses adaptive quantization accuracy based on estimates of posterior uncertainty. Our experimental results demonstrate the importance of taking into account posterior uncertainties, and show that image compression with the proposed algorithm outperforms JPEG over a wide range of bit rates using only a single standard VAE. Further experiments on Bayesian neural word embeddings demonstrate the versatility of the proposed method.

Keywords

Cite

@article{arxiv.2002.08158,
  title  = {Variational Bayesian Quantization},
  author = {Yibo Yang and Robert Bamler and Stephan Mandt},
  journal= {arXiv preprint arXiv:2002.08158},
  year   = {2020}
}

Comments

9 pages + detailed supplement with additional full resolution reconstructed images; ICML 2020 final camera-ready version, title changed to "Variational Bayesian Quantization" following reviewer feedback