Plug-and-Play Methods Provably Converge with Properly Trained Denoisers

Abstract

Plug-and-play (PnP) is a non-convex framework that integrates modern denoising priors, such as BM3D or deep learning-based denoisers, into ADMM or other proximal algorithms. An advantage of PnP is that one can use pre-trained denoisers when there is not sufficient data for end-to-end training. Although PnP has been recently studied extensively with great empirical success, theoretical analysis addressing even the most basic question of convergence has been insufficient. In this paper, we theoretically establish convergence of PnP-FBS and PnP-ADMM, without using diminishing stepsizes, under a certain Lipschitz condition on the denoisers. We then propose real spectral normalization, a technique for training deep learning-based denoisers to satisfy the proposed Lipschitz condition. Finally, we present experimental results validating the theory.

Published in the International Conference on Machine Learning, 2019

Cited by in corpus (30)

• Fixed Point Strategies in Data Science
• Learning to Optimize: A Primer and A Benchmark
• GAP-net for Snapshot Compressive Imaging
• Untrained networks for compressive lensless photography
• Fixed-Point and Objective Convergence of Plug-and-Play Algorithms
• Plug-and-play ISTA converges with kernel denoisers
• On Plug-and-Play Regularization using Linear Denoisers
• SGD-Net: Efficient Model-Based Deep Learning with Theoretical Guarantees
• Solving Inverse Problems by Joint Posterior Maximization with Autoencoding Prior
• Plug-and-Play Quantum Adaptive Denoiser for Deconvolving Poisson Noisy Images
• CoIL: Coordinate-based Internal Learning for Imaging Inverse Problems
• Learning to Optimize in Swarms
• Bayesian imaging using Plug & Play priors: when Langevin meets Tweedie
• Plug-and-Play Algorithms for Video Snapshot Compressive Imaging
• SNIPS: Solving Noisy Inverse Problems Stochastically
• Solving Inverse Problems by Joint Posterior Maximization with a VAE Prior
• Recovery Analysis for Plug-and-Play Priors using the Restricted Eigenvalue Condition
• Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors
• TFPnP: Tuning-free Plug-and-Play Proximal Algorithm with Applications to Inverse Imaging Problems
• Inversion of Integral Models: a Neural Network Approach
• Optimization-Inspired Learning with Architecture Augmentations and Control Mechanisms for Low-Level Vision
• Interpretation of Plug-and-Play (PnP) algorithms from a different angle
• Instabilities in Plug-and-Play (PnP) algorithms from a learned denoiser
• Building Compact and Robust Deep Neural Networks with Toeplitz Matrices
• Bayesian Imaging With Data-Driven Priors Encoded by Neural Networks: Theory, Methods, and Algorithms
• Universal and Flexible Optical Aberration Correction Using Deep-Prior Based Deconvolution
• DWDN: Deep Wiener Deconvolution Network for Non-Blind Image Deblurring
• Joint ptycho-tomography with deep generative priors
• Boosting the Performance of Plug-and-Play Priors via Denoiser Scaling
• Structurally Adaptive Multi-Derivative Regularization for Image Recovery from Sparse Fourier Samples