Learning spectral windowing parameters for regularization using unbiased predictive risk and generalized cross validation techniques for multiple data sets

DOI10.3934/IPI.2023006zbMATH Open1515.65088arXiv2112.12344OpenAlexW4324117842MaRDI QIDQ6042376FDOQ6042376

Authors: Michael J. Byrne, R. A. Renaut 

Publication date: 10 May 2023

Published in: Inverse Problems and Imaging (Search for Journal in Brave)

Abstract: During the inversion of discrete linear systems noise in data can be amplified and result in meaningless solutions. To combat this effect, characteristics of solutions that are considered desirable are mathematically implemented during inversion, which is a process called regularization. The influence of provided prior information is controlled by non-negative regularization parameter(s). There are a number of methods used to select appropriate regularization parameters, as well as a number of methods used for inversion. New methods of unbiased risk estimation and generalized cross validation are derived for finding spectral windowing regularization parameters. These estimators are extended for finding the regularization parameters when multiple data sets with common system matrices are available. It is demonstrated that spectral windowing regularization parameters can be learned from these new estimators applied for multiple data and with multiple windows. The results demonstrate that these modified methods, which do not require the use of true data for learning regularization parameters, are effective and efficient, and perform comparably to a learning method based on estimating the parameters using true data. The theoretical developments are validated for the case of two dimensional image deblurring. The results verify that the obtained estimates of spectral windowing regularization parameters can be used effectively on validation data sets that are separate from the training data, and do not require known data.

Full work available at URL: https://arxiv.org/abs/2112.12344

Recommendations

• Windowed spectral regularization of inverse problems
• Learning regularization parameters for general-form Tikhonov
• Unbiased predictive risk estimation of the Tikhonov regularization parameter: convergence with increasing rank approximations of the singular value decomposition
• On spectral windows in supervised learning from data
• A framework for regularization via operator approximation

zbMATH Keywords

Tikhonov regularizationimagingGSVDmulti-dataspectral windowing

Mathematics Subject Classification ID

Numerical solutions to overdetermined systems, pseudoinverses (65F20) Ill-posedness and regularization problems in numerical linear algebra (65F22)

Cites Work

• An introduction to statistical learning. With applications in R
• IR tools: a MATLAB package of iterative regularization methods and large-scale test problems
• Some Comments on C P
• Generalized Cross-Validation as a Method for Choosing a Good Ridge Parameter
• Title not available (Why is that?)
• The Use of the L-Curve in the Regularization of Discrete Ill-Posed Problems
• Title not available (Why is that?)
• A bilevel optimization approach for parameter learning in variational models
• A Fast Algorithm for Deblurring Models with Neumann Boundary Conditions
• Computational Methods for Inverse Problems
• Title not available (Why is that?)
• Title not available (Why is that?)
• Efficient determination of multiple regularization parameters in a generalized L-curve framework
• Deblurring Images
• Analysis of Discrete Ill-Posed Problems by Means of the L-Curve
• Practical Approximate Solutions to Linear Operator Equations When the Data are Noisy
• Title not available (Why is that?)
• The Discrete Cosine Transform
• Learning from examples as an inverse problem
• Title not available (Why is that?)
• Multi-parameter regularization and its numerical realization
• Multi-parameter regularization techniques for ill-conditioned linear systems
• Multi-parameter Arnoldi-Tikhonov methods
• Towards a Generalized Singular Value Decomposition
• A weighted-GCV method for Lanczos-hybrid regularization
• A Newton root-finding algorithm for estimating the regularization parameter for solving ill-conditioned least squares problems
• Learning regularization functionals a supervised training approach
• A parallel multisplitting solution of the least squares problem
• Designing optimal spectral filters for inverse problems
• Solving inverse problems using data-driven models
• A priori weighting for parameter estimation
• Near-optimal spectral filtering and error estimation for solving ill-posed problems
• Learning regularization parameters for general-form Tikhonov
• Multi-parameter Tikhonov regularization and model function approach to the damped Morozov principle for choosing regularization parameters
• A framework for regularization via operator approximation
• Learning nonlocal regularization operators
• Learning regularization parameters of inverse problems via deep neural networks
• Hybrid and iteratively reweighted regularization by unbiased predictive risk and weighted GCV for projected systems
• Directional multiscale processing of images using wavelets with composite dilations
• Windowed spectral regularization of inverse problems
• Multisplitting for regularized least squares with Krylov subspace recycling.

Cited In (1)

• Windowed spectral regularization of inverse problems

This page was built for publication: Learning spectral windowing parameters for regularization using unbiased predictive risk and generalized cross validation techniques for multiple data sets

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6042376)