Spectral Neural Network Compression via Discrete Fourier Transform: A Post Hoc and Lightweight Approach

Spectral Neural Network Compression via Discrete Fourier Transform: A Post Hoc and Lightweight Approach | 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 Spectral Neural Network Compression via Discrete Fourier Transform: A Post Hoc and Lightweight Approach Sghaier Samia, Nfata Houda This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7254889/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract We introduce a spectral post hoc compression method for neural networks based on Discrete Fourier Transform (DFT) of complex weights. The approach filters low-magnitude frequencies to obtain sparse spectral representations while preserving accuracy. Theoretical results quantify energy preservation and output perturbation. We propose a principled thresholding rule, and demonstrate competitive performance compared to DCT and wavelets. Experiments on MNIST, CIFAR-10 and ResNet show 10–15× compression with negligible loss. Hardware metrics confirm reduced memory usage and improved inference latency. The method is lightweight, requires no retraining , and suits embedded AI. weight compression discrete Fourier Transform frequency-domain filtering complex-valued neural networks deep learning Full Text Additional Declarations No competing interests reported. Supplementary Files WeightCompression22.tex Cite Share Download PDF Status: Posted Version 1 posted 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-7254889","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":495152349,"identity":"2d98a4b4-9687-4c72-8b18-732c18964c05","order_by":0,"name":"Sghaier Samia","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDklEQVRIie2PMUsDMRiGcwRuOunaRe4vBAqZjusP6ZJwUKeI4HqWuPQm6dpJf4EQEZxTPjgHi7ceuFhcLZy4OGRoWuG23DkWzAP5+Ib3yZsg5PEcKdoehrCdDUnsDK71n5VgeTHdK7JXQQfFgqMG2kucxMUL1z/m6nxQ4LKJSJXeF2Bb8mTiUshaqNXN/PlyCGG2HJK37GnNrVJOhXQpSCh9IksuIRohYhWqrRJIcCrxYqtWxpT8DgbfiJHXjFabbgXVQkEU5txOjDTRKa17Wki9VXA61/wBwlEgScZobVtYx1/ihXj8+jQzflvBBzYmHdPqbPPe5In7Yb9Au/FDknXH98zabdwf9ng8nv/GDjEJbl/YPtvBAAAAAElFTkSuQmCC","orcid":"","institution":"Université de Tunis El Manar","correspondingAuthor":true,"prefix":"","firstName":"Sghaier","middleName":"","lastName":"Samia","suffix":""},{"id":495152350,"identity":"982fd4e8-032f-42c2-8ac3-5877261bf463","order_by":1,"name":"Nfata Houda","email":"","orcid":"","institution":"Université de Tunis El Manar","correspondingAuthor":false,"prefix":"","firstName":"Nfata","middleName":"","lastName":"Houda","suffix":""}],"badges":[],"createdAt":"2025-07-30 16:23:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7254889/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7254889/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93640350,"identity":"4f55bf21-4b89-46d0-ab97-ade2f9860616","added_by":"auto","created_at":"2025-10-16 02:33:03","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":408798,"visible":true,"origin":"","legend":"","description":"","filename":"WeightCompression22.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7254889/v1/ef9fd93b87f19626b0e744ae.pdf"},{"id":93640910,"identity":"02741ff5-eda1-4454-a163-0451605cffbe","added_by":"auto","created_at":"2025-10-16 02:41:03","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3484,"visible":true,"origin":"","legend":"","description":"","filename":"d5499b88410440e0b53d2844d806b3a5.json","url":"https://assets-eu.researchsquare.com/files/rs-7254889/v1/14dd31b7f1a05fb81f8a96b6.json"},{"id":93640348,"identity":"ae6950be-9188-403b-aad2-03eb30191e48","added_by":"auto","created_at":"2025-10-16 02:33:03","extension":"tex","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":51691,"visible":true,"origin":"","legend":"","description":"","filename":"WeightCompression22.tex","url":"https://assets-eu.researchsquare.com/files/rs-7254889/v1/3aed3ad952918203e66e608b.tex"},{"id":93641208,"identity":"9210734b-50ff-400f-bc3b-5bde1b92268c","added_by":"auto","created_at":"2025-10-16 02:49:04","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":301055,"visible":true,"origin":"","legend":"","description":"","filename":"WeightCompression22.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7254889/v1_covered_14ce0828-0e43-4e09-af30-41698beb75ee.pdf"},{"id":93640347,"identity":"f526b7ed-f533-4d20-add0-9944005dfbd8","added_by":"auto","created_at":"2025-10-16 02:33:03","extension":"tex","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":51691,"visible":true,"origin":"","legend":"","description":"","filename":"WeightCompression22.tex","url":"https://assets-eu.researchsquare.com/files/rs-7254889/v1/9a76e1b0569c38f6b6a8d1c5.tex"}],"financialInterests":"No competing interests reported.","formattedTitle":"Spectral Neural Network Compression via Discrete Fourier Transform: A Post Hoc and Lightweight Approach","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"weight compression, discrete Fourier Transform, frequency-domain filtering, complex-valued neural networks, deep learning","lastPublishedDoi":"10.21203/rs.3.rs-7254889/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7254889/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"We introduce a spectral post hoc compression method for neural networks based on Discrete Fourier Transform (DFT) of complex weights. The approach filters low-magnitude frequencies to obtain sparse spectral representations while preserving accuracy. Theoretical results quantify energy preservation and output perturbation. We propose a principled thresholding rule, and demonstrate competitive performance compared to DCT and wavelets. Experiments on MNIST, CIFAR-10 and ResNet show 10–15× compression with negligible loss. Hardware metrics confirm reduced memory usage and improved inference latency. The method is lightweight, requires no retraining , and suits embedded AI.","manuscriptTitle":"Spectral Neural Network Compression via Discrete Fourier Transform: A Post Hoc and Lightweight Approach","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-16 02:32:59","doi":"10.21203/rs.3.rs-7254889/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"41d7a9e8-dc6d-4135-bb92-84020a492c61","owner":[],"postedDate":"October 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-16T02:32:59+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-16 02:32:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7254889","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7254889","identity":"rs-7254889","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}