TRUST – Transformer-Driven U-Net for Sparse Target Recovery

Dylan Poppert, Jiayue Li, Mark Foster, Trac D. Tran · 2025

Read TRUST – Transformer-Driven U-Net for Sparse Target Recovery by Dylan Poppert, Jiayue Li, Mark Foster, Trac D. Tran (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-2506-01112.txt.

Description

Recovering a sparse signal from few measurements is a classic hard problem, made harder when the measuring device itself is unknown. A 2025 paper proposes TRUST, a neural design that learns the device and the signal at the same time.

TRUST joins a transformer branch, which reads long-range patterns and guesses where the signal lives, with a U-shaped decoder that refines the rebuild at several scales. Links between the two let fine detail and broad context inform each other. In tests, the hybrid beat classic sparse recovery methods and plain U-shaped networks on standard image quality scores. The setting, learning from a handful of input-output pairs, matches many real sensing problems where the hardware cannot be modeled by hand.

Catalog record

CREATORDylan Poppert, Jiayue Li, Mark Foster, Trac D. Tran
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-01112

Browse the full library · About the Tower of Babel