Saved in:
Bibliographic Details
Main Authors: Sokota, Samuel, Sam, Dylan, de Witt, Christian Schroeder, Compton, Spencer, Foerster, Jakob, Kolter, J. Zico
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.19540
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929365841870848
author Sokota, Samuel
Sam, Dylan
de Witt, Christian Schroeder
Compton, Spencer
Foerster, Jakob
Kolter, J. Zico
author_facet Sokota, Samuel
Sam, Dylan
de Witt, Christian Schroeder
Compton, Spencer
Foerster, Jakob
Kolter, J. Zico
contents Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec .
format Preprint
id arxiv_https___arxiv_org_abs_2405_19540
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Computing Low-Entropy Couplings for Large-Support Distributions
Sokota, Samuel
Sam, Dylan
de Witt, Christian Schroeder
Compton, Spencer
Foerster, Jakob
Kolter, J. Zico
Information Theory
Cryptography and Security
Minimum-entropy coupling (MEC) -- the process of finding a joint distribution with minimum entropy for given marginals -- has applications in areas such as causality and steganography. However, existing algorithms are either computationally intractable for large-support distributions or limited to specific distribution types and sensitive to hyperparameter choices. This work addresses these limitations by unifying a prior family of iterative MEC (IMEC) approaches into a generalized partition-based formalism. From this framework, we derive a novel IMEC algorithm called ARIMEC, capable of handling arbitrary discrete distributions, and introduce a method to make IMEC robust to suboptimal hyperparameter settings. These innovations facilitate the application of IMEC to high-throughput steganography with language models, among other settings. Our codebase is available at https://github.com/ssokota/mec .
title Computing Low-Entropy Couplings for Large-Support Distributions
topic Information Theory
Cryptography and Security
url https://arxiv.org/abs/2405.19540