brendt.wohlberg.net
HomePublications
› Publications
› Software

Cite Details

Yu Sun, Jiaming Liu, Yiran Sun, Brendt Wohlberg and Ulugbek S. Kamilov, "Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors", in Proceedings of the International Conference on Learning Representations (ICLR), (Vienna, Austria), May 2021

Abstract

Regularization by denoising (RED) is a recently developed framework for solving inverse problems by integrating advanced denoisers as image priors. Recent work has shown its state-of-the-art performance when combined with pre-trained deep denoisers. However, current RED algorithms are inadequate for parallel processing on multicore systems. We address this issue by proposing a new asynchronous RED (ASYNC-RED) algorithm that enables asynchronous parallel processing of data, making it significantly faster than its serial counterparts for large-scale inverse problems. The computational complexity of ASYNC-RED is further reduced by using a random subset of measurements at every iteration. We present complete theoretical analysis of the algorithm by establishing its convergence under explicit assumptions on the data-fidelity and the denoiser. We validate ASYNC-RED on image recovery using pre-trained deep denoisers as priors.

BibTeX Entry

@inproceedings{sun-2021-asyncred,
author = {Yu Sun and Jiaming Liu and Yiran Sun and Brendt Wohlberg and Ulugbek S. Kamilov},
title = {{Async-RED}: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors},
year = {2021},
month = May,
urlpdf = {http://openreview.net/pdf?id=9EsrXMzlFQY},
urlhtml = {http://openreview.net/forum?id=9EsrXMzlFQY},
booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)},
address = {Vienna, Austria},
note = {See also arXiv:2010.01446}
}