One-bit Compressed Sensing using Generative Models

Swatantra Kafle, Geethu Joseph, Senior Member, IEEE, Pramod K. Varshney, Life Fellow, IEEE · 2025

Read One-bit Compressed Sensing using Generative Models by Swatantra Kafle, Geethu Joseph, Senior Member, IEEE, Pramod K. Varshney, Life Fellow, IEEE (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-12762.txt.

Description

Swatantra Kafle, Geethu Joseph, and Pramod Varshney present a 2025 paper on recovering signals from very coarse data. In one-bit compressed sensing, each measurement keeps only the sign of a signal value. The puzzle is to rebuild a sparse signal from those signs alone.

Their method uses a trained generative model, a network that maps a small hidden code to realistic signals. To rebuild a signal, the algorithm searches that code space for the output whose signs match the data. Because the model has learned the shape of real signals, it recovers more than sparsity alone allows. The authors prove bounds on accuracy and data needs. Tests on three image datasets beat rival methods and recovered both size and direction.

Catalog record

CREATORSwatantra Kafle, Geethu Joseph, Senior Member, IEEE, Pramod K. Varshney, Life Fellow, IEEE
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-12762

Browse the full library · About the Tower of Babel