Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors
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 a 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.
python 3.6
tensorflow 1.12
scipy 1.2.1
numpy v1.17
matplotlib v3.3.4
It is better to use Conda for installation of all dependecies.
We provide the script
demo_asyncRED.py
to demonstrate the performance of Async-RED with block-diagnal compressive sensing matrix. One can run the code by simply typing
$ python demo_asyncRED.py
To try with different settings, please open the script and follow the instruction inside.
Y. Sun, J. Liu, Y. Sun, B. Wohlberg, and U. S. Kamilov, “Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors,” Proc. Int. Conf. Learn. Represent. (ICLR 2021) (Vienna, Austria, May 4-8).
@inproceedings{
sun2021asyncred,
title={Async-{\{}RED{\}}: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors},
author={Yu Sun and Jiaming Liu and Yiran Sun and Brendt Wohlberg and Ulugbek Kamilov},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/forum?id=9EsrXMzlFQY}
}