Hybrid Projection Methods for Large-scale Inverse Problems with Mixed Gaussian Priors

When solving ill-posed inverse problems, a good choice of the prior is critical for the computation of a reasonable solution. A common approach is to include a Gaussian prior, which is defined by a mean vector and a symmetric and positive definite covariance matrix, and to use iterative projection m...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Cho, Taewon, Chung, Julianne, Jiang, Jiahua
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Cho, Taewon
Chung, Julianne
Jiang, Jiahua
description When solving ill-posed inverse problems, a good choice of the prior is critical for the computation of a reasonable solution. A common approach is to include a Gaussian prior, which is defined by a mean vector and a symmetric and positive definite covariance matrix, and to use iterative projection methods to solve the corresponding regularized problem. However, a main challenge for many of these iterative methods is that the prior covariance matrix must be known and fixed (up to a constant) before starting the solution process. In this paper, we develop hybrid projection methods for inverse problems with mixed Gaussian priors where the prior covariance matrix is a convex combination of matrices and the mixing parameter and the regularization parameter do not need to be known in advance. Such scenarios may arise when data is used to generate a sample prior covariance matrix (e.g., in data assimilation) or when different priors are needed to capture different qualities of the solution. The proposed hybrid methods are based on a mixed Golub-Kahan process, which is an extension of the generalized Golub-Kahan bidiagonalization, and a distinctive feature of the proposed approach is that both the regularization parameter and the weighting parameter for the covariance matrix can be estimated automatically during the iterative process. Furthermore, for problems where training data are available, various data-driven covariance matrices (including those based on learned covariance kernels) can be easily incorporated. Numerical examples from tomographic reconstruction demonstrate the potential for these methods.
doi_str_mv 10.48550/arxiv.2003.13766
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2003_13766</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2003_13766</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-2ec4fb0e4113679d066879996efa80f8defcefbf57c8c77a8f30b0c1b33473413</originalsourceid><addsrcrecordid>eNotz7FOwzAYBGAvDKjwAEz4BRLs2rGdEVXQVkoFQ5mj3_ZvapTGyA6lfXtoYbrhTid9hNxxVkvTNOwB8jEe6jljouZCK3VN3lYnm6Onrzl9oJtiGukGp13yhYaUaQf5HaviYEC6Hg-YC56ndsB9od9x2tFNPKKnS_gqJcL4W8aUyw25CjAUvP3PGdk-P20Xq6p7Wa4Xj10FSqtqjk4Gy1ByLpRuPVPK6LZtFQYwLBiPwWGwodHOOK3BBMEsc9wKIbWQXMzI_d_txdV_5riHfOrPvv7iEz8PI0un</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Hybrid Projection Methods for Large-scale Inverse Problems with Mixed Gaussian Priors</title><source>arXiv.org</source><creator>Cho, Taewon ; Chung, Julianne ; Jiang, Jiahua</creator><creatorcontrib>Cho, Taewon ; Chung, Julianne ; Jiang, Jiahua</creatorcontrib><description>When solving ill-posed inverse problems, a good choice of the prior is critical for the computation of a reasonable solution. A common approach is to include a Gaussian prior, which is defined by a mean vector and a symmetric and positive definite covariance matrix, and to use iterative projection methods to solve the corresponding regularized problem. However, a main challenge for many of these iterative methods is that the prior covariance matrix must be known and fixed (up to a constant) before starting the solution process. In this paper, we develop hybrid projection methods for inverse problems with mixed Gaussian priors where the prior covariance matrix is a convex combination of matrices and the mixing parameter and the regularization parameter do not need to be known in advance. Such scenarios may arise when data is used to generate a sample prior covariance matrix (e.g., in data assimilation) or when different priors are needed to capture different qualities of the solution. The proposed hybrid methods are based on a mixed Golub-Kahan process, which is an extension of the generalized Golub-Kahan bidiagonalization, and a distinctive feature of the proposed approach is that both the regularization parameter and the weighting parameter for the covariance matrix can be estimated automatically during the iterative process. Furthermore, for problems where training data are available, various data-driven covariance matrices (including those based on learned covariance kernels) can be easily incorporated. Numerical examples from tomographic reconstruction demonstrate the potential for these methods.</description><identifier>DOI: 10.48550/arxiv.2003.13766</identifier><language>eng</language><subject>Computer Science - Numerical Analysis ; Mathematics - Numerical Analysis</subject><creationdate>2020-03</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,781,886</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2003.13766$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2003.13766$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Cho, Taewon</creatorcontrib><creatorcontrib>Chung, Julianne</creatorcontrib><creatorcontrib>Jiang, Jiahua</creatorcontrib><title>Hybrid Projection Methods for Large-scale Inverse Problems with Mixed Gaussian Priors</title><description>When solving ill-posed inverse problems, a good choice of the prior is critical for the computation of a reasonable solution. A common approach is to include a Gaussian prior, which is defined by a mean vector and a symmetric and positive definite covariance matrix, and to use iterative projection methods to solve the corresponding regularized problem. However, a main challenge for many of these iterative methods is that the prior covariance matrix must be known and fixed (up to a constant) before starting the solution process. In this paper, we develop hybrid projection methods for inverse problems with mixed Gaussian priors where the prior covariance matrix is a convex combination of matrices and the mixing parameter and the regularization parameter do not need to be known in advance. Such scenarios may arise when data is used to generate a sample prior covariance matrix (e.g., in data assimilation) or when different priors are needed to capture different qualities of the solution. The proposed hybrid methods are based on a mixed Golub-Kahan process, which is an extension of the generalized Golub-Kahan bidiagonalization, and a distinctive feature of the proposed approach is that both the regularization parameter and the weighting parameter for the covariance matrix can be estimated automatically during the iterative process. Furthermore, for problems where training data are available, various data-driven covariance matrices (including those based on learned covariance kernels) can be easily incorporated. Numerical examples from tomographic reconstruction demonstrate the potential for these methods.</description><subject>Computer Science - Numerical Analysis</subject><subject>Mathematics - Numerical Analysis</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz7FOwzAYBGAvDKjwAEz4BRLs2rGdEVXQVkoFQ5mj3_ZvapTGyA6lfXtoYbrhTid9hNxxVkvTNOwB8jEe6jljouZCK3VN3lYnm6Onrzl9oJtiGukGp13yhYaUaQf5HaviYEC6Hg-YC56ndsB9od9x2tFNPKKnS_gqJcL4W8aUyw25CjAUvP3PGdk-P20Xq6p7Wa4Xj10FSqtqjk4Gy1ByLpRuPVPK6LZtFQYwLBiPwWGwodHOOK3BBMEsc9wKIbWQXMzI_d_txdV_5riHfOrPvv7iEz8PI0un</recordid><startdate>20200330</startdate><enddate>20200330</enddate><creator>Cho, Taewon</creator><creator>Chung, Julianne</creator><creator>Jiang, Jiahua</creator><scope>AKY</scope><scope>AKZ</scope><scope>GOX</scope></search><sort><creationdate>20200330</creationdate><title>Hybrid Projection Methods for Large-scale Inverse Problems with Mixed Gaussian Priors</title><author>Cho, Taewon ; Chung, Julianne ; Jiang, Jiahua</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-2ec4fb0e4113679d066879996efa80f8defcefbf57c8c77a8f30b0c1b33473413</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Numerical Analysis</topic><topic>Mathematics - Numerical Analysis</topic><toplevel>online_resources</toplevel><creatorcontrib>Cho, Taewon</creatorcontrib><creatorcontrib>Chung, Julianne</creatorcontrib><creatorcontrib>Jiang, Jiahua</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Mathematics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Cho, Taewon</au><au>Chung, Julianne</au><au>Jiang, Jiahua</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Hybrid Projection Methods for Large-scale Inverse Problems with Mixed Gaussian Priors</atitle><date>2020-03-30</date><risdate>2020</risdate><abstract>When solving ill-posed inverse problems, a good choice of the prior is critical for the computation of a reasonable solution. A common approach is to include a Gaussian prior, which is defined by a mean vector and a symmetric and positive definite covariance matrix, and to use iterative projection methods to solve the corresponding regularized problem. However, a main challenge for many of these iterative methods is that the prior covariance matrix must be known and fixed (up to a constant) before starting the solution process. In this paper, we develop hybrid projection methods for inverse problems with mixed Gaussian priors where the prior covariance matrix is a convex combination of matrices and the mixing parameter and the regularization parameter do not need to be known in advance. Such scenarios may arise when data is used to generate a sample prior covariance matrix (e.g., in data assimilation) or when different priors are needed to capture different qualities of the solution. The proposed hybrid methods are based on a mixed Golub-Kahan process, which is an extension of the generalized Golub-Kahan bidiagonalization, and a distinctive feature of the proposed approach is that both the regularization parameter and the weighting parameter for the covariance matrix can be estimated automatically during the iterative process. Furthermore, for problems where training data are available, various data-driven covariance matrices (including those based on learned covariance kernels) can be easily incorporated. Numerical examples from tomographic reconstruction demonstrate the potential for these methods.</abstract><doi>10.48550/arxiv.2003.13766</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2003.13766
ispartof
issn
language eng
recordid cdi_arxiv_primary_2003_13766
source arXiv.org
subjects Computer Science - Numerical Analysis
Mathematics - Numerical Analysis
title Hybrid Projection Methods for Large-scale Inverse Problems with Mixed Gaussian Priors
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-18T13%3A09%3A08IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Hybrid%20Projection%20Methods%20for%20Large-scale%20Inverse%20Problems%20with%20Mixed%20Gaussian%20Priors&rft.au=Cho,%20Taewon&rft.date=2020-03-30&rft_id=info:doi/10.48550/arxiv.2003.13766&rft_dat=%3Carxiv_GOX%3E2003_13766%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true