Taming Score-Based Denoisers in ADMM: A Convergent Plug-and-Play Framework

Rajesh Shrestha, Xiao Fu · 2026

Read Taming Score-Based Denoisers in ADMM: A Convergent Plug-and-Play Framework by Rajesh Shrestha, Xiao Fu (2026) 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-2603-10281.txt.

Description

Rajesh Shrestha and Xiao Fu tackle a hard fit: plugging score-based denoisers into ADMM for inverse problems without losing convergence. Two snags drive the work, a mismatch between training noise and ADMM iterates, plus thin theory for the pairing. Their AC-DC denoiser adds noise for auto-correction, steers iterates with conditional Langevin steps, then denoises by score. The authors prove fixed-point and adaptive-step convergence results under stated conditions, and experiments show better solution quality than a range of baselines.

Catalog record

CREATORRajesh Shrestha, Xiao Fu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE129239
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.10281

Browse the full library · About the Tower of Babel