QuATON: Quantization Aware Training of Optical Neurons

QuATON: Quantization Aware Training of Optical Neurons | 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 Article QuATON: Quantization Aware Training of Optical Neurons Dushan Wadduwage, Hasindu Kariyawasam, Ramith Hettiarachchi, Quansan Yang, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4076842/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 Optical processors, built with "optical neurons", can efficiently perform high-dimensional linear operations at the speed of light. Thus they are a promising avenue to accelerate large-scale linear computations. With the current advances in micro-fabrication, such optical processors can now be 3D fabricated, but with a limited precision. This limitation translates to quantization of learnable parameters in optical neurons, and should be handled during the design of the optical processor in order to avoid a model mismatch. Specifically, optical neurons should be trained or designed within the physical-constraints at a predefined quantized precision level. To address this critical issues we propose a physics-informed quantization-aware training framework. Our approach accounts for physical constraints during the training process, leading to robust designs. We demonstrate that our approach can design state of the art optical processors using diffractive networks for multiple physics based tasks despite quantized learnable parameters. We thus lay the foundation upon which improved optical processors may be 3D fabricated in the future. Physical sciences/Optics and photonics/Optical techniques/Lithography Physical sciences/Optics and photonics/Other photonics/Biophotonics Full Text Additional Declarations (Not answered) 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-4076842","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":283439320,"identity":"0de5c7f3-9632-4a63-9f8e-39fdd2036644","order_by":0,"name":"Dushan Wadduwage","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDUlEQVRIiWNgGAWjYHACAwkwxd74AEgegIjxMDAwNhDUwnPYgFQtEslEajE4fnjjbd42O7t+yceMH38w3JE3n5H87MMbBhvZDQdwaDmTVmzN25acPHN2MrM0D8Mzwzk30oxnzmFIM8alxexAjpk07zbmZIPb+QekGRgOM87gOWDMzMNwOBGnlvNvQFrqk+1vHmb++YPhsP0MnuOfgVr+49ZyA2zLYTsDCWY2CZDhM9h7QLYcwKnF/sazYsu5/44nSJxJZrPmMTicDNRSzDjHINl4Jg4tkv3JG2+8OVNtz99+mPnmj4rDtjOY2TczvKmwk+3DoQUGEhvAlAGMb4BLIZIDCSsZBaNgFIyCEQsACz1fZn+fRPIAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-0689-9454","institution":"Harvard University","correspondingAuthor":true,"prefix":"","firstName":"Dushan","middleName":"","lastName":"Wadduwage","suffix":""},{"id":283439321,"identity":"1333729d-0eda-48b2-99d7-ab37471ee7d0","order_by":1,"name":"Hasindu Kariyawasam","email":"","orcid":"","institution":"Harvard University","correspondingAuthor":false,"prefix":"","firstName":"Hasindu","middleName":"","lastName":"Kariyawasam","suffix":""},{"id":283439322,"identity":"c3f9f787-71c0-4823-a1e5-aeefefad0f89","order_by":2,"name":"Ramith Hettiarachchi","email":"","orcid":"","institution":"Harvard University","correspondingAuthor":false,"prefix":"","firstName":"Ramith","middleName":"","lastName":"Hettiarachchi","suffix":""},{"id":283439323,"identity":"370ef13c-07f2-42af-b791-1bba228b8984","order_by":3,"name":"Quansan Yang","email":"","orcid":"","institution":"Massachussetts Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Quansan","middleName":"","lastName":"Yang","suffix":""},{"id":283439324,"identity":"4cb211ad-c64f-400e-b713-021b8992f4cf","order_by":4,"name":"Alex Matlock","email":"","orcid":"https://orcid.org/0000-0002-4337-2715","institution":"Massachusetts Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Alex","middleName":"","lastName":"Matlock","suffix":""},{"id":283439325,"identity":"13e80a3c-fc6c-42c6-adca-89186d392ede","order_by":5,"name":"Takahiro Nambara","email":"","orcid":"","institution":"Massachussetts Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Takahiro","middleName":"","lastName":"Nambara","suffix":""},{"id":283439326,"identity":"6af58008-95e8-4c5a-8e44-6da7457dcc76","order_by":6,"name":"Hiroyuki Kusaka","email":"","orcid":"","institution":"Massachussetts Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Hiroyuki","middleName":"","lastName":"Kusaka","suffix":""},{"id":283439327,"identity":"96a8d34f-4002-4139-a998-cf46d51650ac","order_by":7,"name":"Yuichiro Kunai","email":"","orcid":"","institution":"Massachussetts Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Yuichiro","middleName":"","lastName":"Kunai","suffix":""},{"id":283439328,"identity":"0847e08b-f826-4427-aec6-1428749618c4","order_by":8,"name":"Peter So","email":"","orcid":"https://orcid.org/0000-0003-4698-6488","institution":"Dept of Mechanical Engineering","correspondingAuthor":false,"prefix":"","firstName":"Peter","middleName":"","lastName":"So","suffix":""},{"id":283439329,"identity":"b4382e4d-1a5c-4561-b9fd-39c55eee8080","order_by":9,"name":"Edward Boyden","email":"","orcid":"","institution":"Massachussetts Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Edward","middleName":"","lastName":"Boyden","suffix":""}],"badges":[],"createdAt":"2024-03-11 17:01:42","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4076842/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4076842/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":55281608,"identity":"0215e82b-49ce-4af2-a14a-bd2e494a9de8","added_by":"auto","created_at":"2024-04-25 06:58:29","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8794349,"visible":true,"origin":"","legend":"","description":"","filename":"manuscriptKariyawasametal.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4076842/v1_covered_92ee139e-f146-4e16-af85-c57f894ba246.pdf"}],"financialInterests":"(Not answered)","formattedTitle":"QuATON: Quantization Aware Training of Optical Neurons","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"","lastPublishedDoi":"10.21203/rs.3.rs-4076842/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4076842/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Optical processors, built with \"optical neurons\", can efficiently perform high-dimensional linear operations at the speed of light. Thus they are a promising avenue to accelerate large-scale linear computations. With the current advances in micro-fabrication, such optical processors can now be 3D fabricated, but with a limited precision. This limitation translates to quantization of learnable parameters in optical neurons, and should be handled during the design of the optical processor in order to avoid a model mismatch. Specifically, optical neurons should be trained or designed within the physical-constraints at a predefined quantized precision level. To address this critical issues we propose a physics-informed quantization-aware training framework. Our approach accounts for physical constraints during the training process, leading to robust designs. We demonstrate that our approach can design state of the art optical processors using diffractive networks for multiple physics based tasks despite quantized learnable parameters. We thus lay the foundation upon which improved optical processors may be 3D fabricated in the future.","manuscriptTitle":"QuATON: Quantization Aware Training of Optical Neurons","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-27 17:23:32","doi":"10.21203/rs.3.rs-4076842/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":"bc870736-4ca1-40e8-9565-a9a59636892a","owner":[],"postedDate":"March 27th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":29835470,"name":"Physical sciences/Optics and photonics/Optical techniques/Lithography"},{"id":29835471,"name":"Physical sciences/Optics and photonics/Other photonics/Biophotonics"}],"tags":[],"updatedAt":"2024-04-25T06:50:18+00:00","versionOfRecord":[],"versionCreatedAt":"2024-03-27 17:23:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4076842","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4076842","identity":"rs-4076842","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}