Efficient Model-based Deep Learning via Network Pruning and Fine-Tuning

Efficient Model-based Deep Learning via Network Pruning and Fine-Tuning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Efficient Model-based Deep Learning via Network Pruning and Fine-Tuning Chicago Y. Park, Weijie Gan, Zihao Zou, Yuyang Hu, Zhixin Sun, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5286110/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 07 Apr, 2025 Read the published version in Journal of Mathematical Imaging and Vision → Version 1 posted 9 You are reading this latest preprint version Abstract Model-based deep learning (MBDL) is a powerful methodology for designing deep models to solve imaging inverse problems. MBDL networks can be seen as iterative algorithms that estimate the desired image using a physical measurement model and a learned image prior specified using a convolutional neural net (CNNs). The iterative nature of MBDL networks increases the test-time computational complexity, which limits their applicability in certain large-scale applications. Here we make two contributions to address this issue: First, we show how structured pruning can be adopted to reduce the number of parameters in MBDL networks. Second, we present three methods to fine-tune the pruned MBDL networks to mitigate potential performance loss. Each fine-tuning strategy has a unique benefit that depends on the presence of a pre-trained model and a high-quality ground truth. We show that our pruning and fine-tuning approach can accelerate image reconstruction using popular deep equilibrium learning (DEQ) and deep unfolding (DU) methods by 50% and 32%, respectively, with nearly no performance loss. This work thus offers a step forward for solving inverse problems by showing the potential of pruning to improve the scalability of MBDL. Computational imaging inverse problem model-based deep learning network pruning Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 07 Apr, 2025 Read the published version in Journal of Mathematical Imaging and Vision → Version 1 posted Editorial decision: Revision requested 17 Feb, 2025 Reviews received at journal 11 Feb, 2025 Reviewers agreed at journal 27 Jan, 2025 Reviews received at journal 28 Dec, 2024 Reviewers agreed at journal 30 Oct, 2024 Reviewers invited by journal 28 Oct, 2024 Editor assigned by journal 24 Oct, 2024 Submission checks completed at journal 18 Oct, 2024 First submitted to journal 17 Oct, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5286110","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":371360414,"identity":"27ec4a6a-4840-49b4-9ebc-34354fd14bf2","order_by":0,"name":"Chicago Y. Park","email":"","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":false,"prefix":"","firstName":"Chicago","middleName":"Y.","lastName":"Park","suffix":""},{"id":371360415,"identity":"56928d3b-0440-4efe-ac12-d78355b3c256","order_by":1,"name":"Weijie Gan","email":"","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":false,"prefix":"","firstName":"Weijie","middleName":"","lastName":"Gan","suffix":""},{"id":371360416,"identity":"60a40d38-fc22-4014-84df-10bd747493ac","order_by":2,"name":"Zihao Zou","email":"","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":false,"prefix":"","firstName":"Zihao","middleName":"","lastName":"Zou","suffix":""},{"id":371360417,"identity":"414a624f-918b-463f-b663-1715c6970082","order_by":3,"name":"Yuyang Hu","email":"","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":false,"prefix":"","firstName":"Yuyang","middleName":"","lastName":"Hu","suffix":""},{"id":371360418,"identity":"b89c5c6f-bf00-4f99-af97-2aa405848586","order_by":4,"name":"Zhixin Sun","email":"","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":false,"prefix":"","firstName":"Zhixin","middleName":"","lastName":"Sun","suffix":""},{"id":371360419,"identity":"20439126-7c28-47fc-ad8e-b67b12909266","order_by":5,"name":"Ulugbek S. Kamilov","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAw0lEQVRIiWNgGAWjYBACCSA+8ICBQQ7MAzIYG4jSksDAYMwAUp1ArBYGoMrEBqK1SLafTjyQUHE4fcON9OcPEhhsZDccIKBFmid3w4GEM4dzN9zIMQTakmZMUIscA1BLYlta7rYbOSCHHU4krIX/LVhLutmN9IdALf8Ja5GWANtik2B2IwHksAOEtUjOeAvyi43h/jNvDGckGCQbzySkReJ87uYPHyok5CXb0x8AGXayfYS0oAED0pSPglEwCkbBKMABAOaFTbq8FWUiAAAAAElFTkSuQmCC","orcid":"","institution":"Washington University in St. Louis","correspondingAuthor":true,"prefix":"","firstName":"Ulugbek","middleName":"S.","lastName":"Kamilov","suffix":""}],"badges":[],"createdAt":"2024-10-18 03:38:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5286110/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5286110/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10851-025-01241-1","type":"published","date":"2025-04-07T16:05:46+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":80558675,"identity":"0423457e-daa8-42b2-927e-94e5b41aae3c","added_by":"auto","created_at":"2025-04-14 16:15:59","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1044912,"visible":true,"origin":"","legend":"","description":"","filename":"jmiv2024.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5286110/v1_covered_463c63ba-1d3f-4db9-9e16-647248921551.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Efficient Model-based Deep Learning via Network Pruning and Fine-Tuning","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-mathematical-imaging-and-vision","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jmiv","sideBox":"Learn more about [Journal of Mathematical Imaging and Vision](http://link.springer.com/journal/10851)","snPcode":"10851","submissionUrl":"https://submission.nature.com/new-submission/10851/3","title":"Journal of Mathematical Imaging and Vision","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Computational imaging, inverse problem, model-based deep learning, network pruning","lastPublishedDoi":"10.21203/rs.3.rs-5286110/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5286110/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Model-based deep learning (MBDL) is a powerful methodology for designing deep models to solve imaging inverse problems. MBDL networks can be seen as iterative algorithms that estimate the desired image using a physical measurement model and a learned image prior specified using a convolutional neural net (CNNs). The iterative nature of MBDL networks increases the test-time computational complexity, which limits their applicability in certain large-scale applications. Here we make two contributions to address this issue: First, we show how structured pruning can be adopted to reduce the number of parameters in MBDL networks. Second, we present three methods to fine-tune the pruned MBDL networks to mitigate potential performance loss. Each fine-tuning strategy has a unique benefit that depends on the presence of a pre-trained model and a high-quality ground truth. We show that our pruning and fine-tuning approach can accelerate image reconstruction using popular deep equilibrium learning (DEQ) and deep unfolding (DU) methods by 50% and 32%, respectively, with nearly no performance loss. This work thus offers a step forward for solving inverse problems by showing the potential of pruning to improve the scalability of MBDL.","manuscriptTitle":"Efficient Model-based Deep Learning via Network Pruning and Fine-Tuning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-10-29 04:09:19","doi":"10.21203/rs.3.rs-5286110/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-02-17T14:22:10+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-02-11T11:32:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"243386160780350271346533426526340497803","date":"2025-01-27T10:46:15+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-28T14:19:50+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"52085509330654639823328906961169605759","date":"2024-10-30T16:29:40+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-10-28T16:05:07+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-10-24T12:06:07+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-10-18T05:05:17+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Mathematical Imaging and Vision","date":"2024-10-18T03:28:19+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-mathematical-imaging-and-vision","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jmiv","sideBox":"Learn more about [Journal of Mathematical Imaging and Vision](http://link.springer.com/journal/10851)","snPcode":"10851","submissionUrl":"https://submission.nature.com/new-submission/10851/3","title":"Journal of Mathematical Imaging and Vision","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"8e62b576-6ecd-448c-9814-49654b5fff6d","owner":[],"postedDate":"October 29th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-04-14T16:11:06+00:00","versionOfRecord":{"articleIdentity":"rs-5286110","link":"https://doi.org/10.1007/s10851-025-01241-1","journal":{"identity":"journal-of-mathematical-imaging-and-vision","isVorOnly":false,"title":"Journal of Mathematical Imaging and Vision"},"publishedOn":"2025-04-07 16:05:46","publishedOnDateReadable":"April 7th, 2025"},"versionCreatedAt":"2024-10-29 04:09:19","video":"","vorDoi":"10.1007/s10851-025-01241-1","vorDoiUrl":"https://doi.org/10.1007/s10851-025-01241-1","workflowStages":[]},"version":"v1","identity":"rs-5286110","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5286110","identity":"rs-5286110","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}