Data-Free Pruning of CNN Using Kernel Similarity

Data-Free Pruning of CNN Using Kernel Similarity | 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 Data-Free Pruning of CNN Using Kernel Similarity Xinwang Chen, Fengrui Ji, Renxin Chu, Baolin Liu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4919297/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 13 Mar, 2025 Read the published version in Multimedia Systems → Version 1 posted 14 You are reading this latest preprint version Abstract Channel pruning can effectively compress Convolutional Neural Networks (CNNs) for deployment on edge devices. Most existing pruning methods are data-driven, relying heavily on datasets and necessitating fine-tuning the pruned models for several epochs. However, data privacy protection increases the difficulty of getting a dataset, making data inaccessible in some scenarios. Inaccessible datasets lead to current pruning methods infeasible. To solve this issue, we propose a data-free CNN pruning method that does not require data. It involves filter reconstruction and feature reconstruction. To reduce kernels in each filter, we group the kernels in each filter based on the similarity of kernels and calculate a representative kernel for each group to reconstruct the filters. During inference, we conduct feature reconstruction to match input channels of the reconstructed filter so as to satisfy the operational criteria of convolutional neural networks. We validate the effectiveness of our method through extensive experiments using ResNet, MobileNet, and VGG on CIFAR-10 and ImageNet datasets. For ResNet-50, we obtain FLOPs reduction of 56.2% with only Top-1 accuracy reduction of 0.52% on ImageNet. Channel Pruning Data-free Method Kernel Similarity Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 13 Mar, 2025 Read the published version in Multimedia Systems → Version 1 posted Editorial decision: Revision requested 02 Dec, 2024 Reviews received at journal 23 Nov, 2024 Reviews received at journal 20 Nov, 2024 Reviewers agreed at journal 15 Nov, 2024 Reviewers agreed at journal 31 Oct, 2024 Reviews received at journal 16 Oct, 2024 Reviewers agreed at journal 26 Sep, 2024 Reviews received at journal 23 Sep, 2024 Reviewers agreed at journal 23 Sep, 2024 Reviewers agreed at journal 23 Sep, 2024 Reviewers invited by journal 23 Sep, 2024 Editor assigned by journal 17 Sep, 2024 Submission checks completed at journal 16 Aug, 2024 First submitted to journal 15 Aug, 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-4919297","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":385396370,"identity":"d67c0da8-1cbf-4559-91ec-46984a5fb1fe","order_by":0,"name":"Xinwang Chen","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Xinwang","middleName":"","lastName":"Chen","suffix":""},{"id":385396371,"identity":"ed411c24-b9e3-43cf-9ab3-788cf02e5684","order_by":1,"name":"Fengrui Ji","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Fengrui","middleName":"","lastName":"Ji","suffix":""},{"id":385396372,"identity":"69320917-deee-4e9b-8c48-d680780d5999","order_by":2,"name":"Renxin Chu","email":"","orcid":"","institution":"Harvard Medical School","correspondingAuthor":false,"prefix":"","firstName":"Renxin","middleName":"","lastName":"Chu","suffix":""},{"id":385396373,"identity":"663839d9-8a33-4c30-b85e-30fc65f1689f","order_by":3,"name":"Baolin Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIiWNgGAWjYBACPmYGNiBlwcPA3sDAwAjEDAcIaGGDaJHgYeABKj1IlBYGiBYgSiBWCzuP2YMfFRIy5pJvjD9/3MEgx3cjgfFzAV6H8Zgb9pyR4LGcnWMmcfAMg7HkjQRm6Rn4tZhJ8LZJ8BjczjFjONjGkLjhRgJQkIAWyb8gLTfPGH8AaqknSos02JYbPAYSQC0JBoS1sJUbywD9YnAmrUzi7BkJw5lnHjZL49PCz39428M3FTb2BscPb/5QucNGnu948sHP+LSgA2AEQdLAKBgFo2AUjAJKAABqnELc9gTJUwAAAABJRU5ErkJggg==","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":true,"prefix":"","firstName":"Baolin","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2024-08-15 12:30:40","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4919297/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4919297/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00530-025-01743-3","type":"published","date":"2025-03-13T15:58:52+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":78689668,"identity":"275084c0-1eae-4750-a6ec-7470e5313ee6","added_by":"auto","created_at":"2025-03-17 16:12:55","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4186104,"visible":true,"origin":"","legend":"","description":"","filename":"DataFreePruningofCNNUsingKernelSimilarity.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4919297/v1_covered_33675924-9c93-4253-8b48-a05810bb8d2c.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Data-Free Pruning of CNN Using Kernel Similarity","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":"multimedia-systems","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mmsj","sideBox":"Learn more about [Multimedia Systems](http://link.springer.com/journal/530)","snPcode":"530","submissionUrl":"https://submission.nature.com/new-submission/530/3","title":"Multimedia Systems","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Channel Pruning, Data-free Method, Kernel Similarity","lastPublishedDoi":"10.21203/rs.3.rs-4919297/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4919297/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Channel pruning can effectively compress Convolutional Neural Networks (CNNs) for deployment on edge devices. Most existing pruning methods are data-driven, relying heavily on datasets and necessitating fine-tuning the pruned models for several epochs. However, data privacy protection increases the difficulty of getting a dataset, making data inaccessible in some scenarios. Inaccessible datasets lead to current pruning methods infeasible. To solve this issue, we propose a data-free CNN pruning method that does not require data. It involves filter reconstruction and feature reconstruction. To reduce kernels in each filter, we group the kernels in each filter based on the similarity of kernels and calculate a representative kernel for each group to reconstruct the filters. During inference, we conduct feature reconstruction to match input channels of the reconstructed filter so as to satisfy the operational criteria of convolutional neural networks. We validate the effectiveness of our method through extensive experiments using ResNet, MobileNet, and VGG on CIFAR-10 and ImageNet datasets. For ResNet-50, we obtain FLOPs reduction of 56.2% with only Top-1 accuracy reduction of 0.52% on ImageNet.","manuscriptTitle":"Data-Free Pruning of CNN Using Kernel Similarity","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-18 01:41:00","doi":"10.21203/rs.3.rs-4919297/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-12-03T00:26:17+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-11-23T08:11:01+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-11-20T06:38:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"316790028773715136187024127614402116271","date":"2024-11-15T06:49:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"54018840258243784424726485036718475364","date":"2024-11-01T02:03:53+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-10-16T23:05:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"204278849595181905324639989434859138421","date":"2024-09-26T05:52:44+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-09-23T15:51:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"203660236318060758591553501029542179373","date":"2024-09-23T14:15:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"238633606360937599495219836091602468316","date":"2024-09-23T08:30:17+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-09-23T06:06:57+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-09-17T09:53:29+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-08-16T13:41:32+00:00","index":"","fulltext":""},{"type":"submitted","content":"Multimedia Systems","date":"2024-08-15T12:29:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"multimedia-systems","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mmsj","sideBox":"Learn more about [Multimedia Systems](http://link.springer.com/journal/530)","snPcode":"530","submissionUrl":"https://submission.nature.com/new-submission/530/3","title":"Multimedia Systems","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"a2e86b22-8380-4a25-845a-ff85f1ff9b43","owner":[],"postedDate":"December 18th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-03-17T16:08:40+00:00","versionOfRecord":{"articleIdentity":"rs-4919297","link":"https://doi.org/10.1007/s00530-025-01743-3","journal":{"identity":"multimedia-systems","isVorOnly":false,"title":"Multimedia Systems"},"publishedOn":"2025-03-13 15:58:52","publishedOnDateReadable":"March 13th, 2025"},"versionCreatedAt":"2024-12-18 01:41:00","video":"","vorDoi":"10.1007/s00530-025-01743-3","vorDoiUrl":"https://doi.org/10.1007/s00530-025-01743-3","workflowStages":[]},"version":"v1","identity":"rs-4919297","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4919297","identity":"rs-4919297","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}