Fetching the paper…

Reading the bibliography…

2021

On the advantages of stochastic encoders

Theis, Lucas, Agustsson, Eirikur

Original
PDF
Post on X

Understand

Stochastic encoders have been used in rate-distortion theory and neural compression because they can be easier to handle.

However, in performance comparisons with deterministic encoders they often do worse, suggesting that noise in the encoding process may generally be a bad idea.
It is poorly understood if and when stochastic encoders do better than deterministic encoders.
In this paper we provide one illustrative example which shows that stochastic encoders can significantly outperform the best deterministic encoders.

On the advantages of stochastic encoders · Around

Built on

Efficient scalar quantization of exponential and Laplacian random variables
G. J. Sullivan · 1996
Earlier work this paper cites.
Neural Networks Optimally Compress the Sawbridge
Original
A. B. Wagner and J. Ballé · 2011
Earlier work this paper cites.
The perception-distortion tradeoff
Y. Blau and T. Michaeli · 2018
Earlier work this paper cites.

Similar

Deep Generative Models for Distribution-Preserving Lossy Compression
M. Tschannen, E. Agustsson, and M. Lucic · 2018
Cited alongside, same era.
Rethinking lossy compression: The rate-distortion-perception tradeoff
Y. Blau and T. Michaeli · 2019
Cited alongside, same era.
Minimal Random Code Learning: Getting Bits Back from Compressed Model Parameters
M. Havasi, R. Peharz, and J. M. Hernández-Lobato · 2019
Cited alongside, same era.

Then

Universally Quantized Neural Compression
E. Agustsson and L. Theis · 2020
Later among the works it cites.
Nonlinear transform coding
J. Ballé, P. A. Chou, D. Minnen, S. Singh, N. Johnston, E. Agustsson, S. J. Hwang, and G. Toderici · 2021
Closest in time.
On universal quantization
J. Ziv · 2021
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…