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.

arXiv commentPublished in the International Conference on Machine Learning, 2019

Cited by in corpus (42)

• RARE: Image Reconstruction using Deep Priors Learned without Ground Truth
• Deep Plug-and-play Prior for Low-rank Tensor Completion
• 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
• SIMBA: Scalable Inversion in Optical Tomography using Deep Denoising Priors
• 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
• CPGD: Cadzow Plug-and-Play Gradient Descent for Generalised FRI
• SGD-Net: Efficient Model-Based Deep Learning with Theoretical Guarantees
• Investigating Task-driven Latent Feasibility for Nonconvex Image Modeling
• 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
• Plug-and-Play Algorithms for Large-scale Snapshot Compressive Imaging
• Tuning-free Plug-and-Play Proximal Algorithm for Inverse Imaging Problems
• Learning to Optimize in Swarms
• Bayesian imaging using Plug & Play priors: when Langevin meets Tweedie
• Fantastic Four: Differentiable Bounds on Singular Values of Convolution Layers
• Plug-and-Play Algorithms for Video Snapshot Compressive Imaging
• Solving Inverse Problems by Joint Posterior Maximization with a VAE Prior
• SNIPS: Solving Noisy Inverse Problems Stochastically
• Async-RED: A Provably Convergent Asynchronous Block Parallel Stochastic Method using Deep Denoising Priors
• Recovery Analysis for Plug-and-Play Priors using the Restricted Eigenvalue Condition
• Momentum-Net for Low-Dose CT Image Reconstruction
• TFPnP: Tuning-free Plug-and-Play Proximal Algorithm with Applications to Inverse Imaging Problems
• Scalable Plug-and-Play ADMM with Convergence Guarantees
• Optimization-Inspired Learning with Architecture Augmentations and Control Mechanisms for Low-Level Vision
• Inversion of Integral Models: a Neural Network Approach
• Bayesian Imaging With Data-Driven Priors Encoded by Neural Networks: Theory, Methods, and Algorithms
• Plug-and-Play Priors for Reconstruction-based Placental Image Registration
• Joint ptycho-tomography with deep generative priors
• DWDN: Deep Wiener Deconvolution Network for Non-Blind Image Deblurring
• Universal and Flexible Optical Aberration Correction Using Deep-Prior Based Deconvolution
• 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
• Online Regularization by Denoising with Applications to Phase Retrieval
• Boosting the Performance of Plug-and-Play Priors via Denoiser Scaling
• Structurally Adaptive Multi-Derivative Regularization for Image Recovery from Sparse Fourier Samples