High-speed feature extraction with integrated microwave neurons

High-speed feature extraction with integrated microwave 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 High-speed feature extraction with integrated microwave neurons BAL GOVIND, Pablo RAIGOZA, ALYSSA APSEL This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8356393/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Intelligence at high data rates depends on extracting meaning the instant information arrives. However, while real visual, audio and radio signals are fundamentally analog, most computers still fully digitize them before extracting features, which constrains processing speed to the pace of a digital clock and adds latency and power overhead from analog-to-digital conversion. We find that, instead, an ensemble of coupled waveguides could extract spectral features of incoming signals instantaneously through interactions between frequency modes across tens of gigahertz. This milliwatt-scale Microwave Neural Network (MNN) uses the incoming signals themselves to reconfigure coupling between waveguides, reshaping its spectrum in real time. By instantly expressing each token’s features across many frequencies, the MNN makes the relationships between successive tokens easier to detect, enabling faster decision-making — without digital preprocessing. Also, when the waveguides are driven directly by gigabit-per-second bitstreams, such as 8-bit pixel data, the MNN produces probabilistic-bits whose bias reflects the input pattern. This acts like an analog form of dithering with lower overhead, where controlled randomness preserves fine detail using fewer bits. As a result, the readout can sample at only one-eigth of the bit rate while still retaining the distinguishing features of the original pattern. This could let satellites downlink richer imagery within tight bandwidth limits and enable radar sensors to classify threats from radio-frequency signals as they arrive—embedding context directly into clockless front-end circuitry. Physical sciences/Mathematics and computing/Computer science Physical sciences/Engineering/Electrical and electronic engineering Physical sciences/Physics/Electronics, photonics and device physics/Electronic and spintronic devices Figures Figure 1 Figure 2 Figure 3 Figure 3 Figure 4 Figure 5 Figure 6 Full Text Additional Declarations Yes there is potential Competing Interest. B.G. and A.A. have filed a provisional US patent application (number 63/742,208) based on the frontend processor presented in this article. P.R. declares no competing interests. Supplementary Files EX1.pdf Extended Data Fig 1 EX4.pdf Extended Data Fig 4 EX2.pdf Extended Data Fig 2 EX3.pdf Extended Data Fig 3 EX4.pdf Extended Data Fig 4 EX5compressed.pdf Extended Data Fig 5 EX4.pdf SupplementHighspeedfeatureextraction.pdf Supplementary Information for High-speed feature extraction with integrated microwave neurons Cite Share Download PDF Status: Under Review 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-8356393","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":569755229,"identity":"bda88963-944a-4681-9829-774674c6cf5b","order_by":0,"name":"BAL GOVIND","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzUlEQVRIiWNgGAWjYBACAwb24z8/VIDZjAceAIkGwlp4EqQlzkA4BxKI08JgIMHbRooWc/4FCQaS8+zkzGc3HwBqsZHdcICAFssZDw8kFG5LNpa5cywBqCXNmKAWgxsHEg5IbjuQOEMixwCo5XAiMVoMG3jngLTkfwBq+U+ElvMNxgy8DWBbQN4/QFiL5QyeNGaJY8nGEhJpQIcZJBvPJKTFnP/4McYPNXZyEhLJDx98qLCT7SOkhUEiAcWdhJSDAD9BQ0fBKBgFo2DEAwCo+0pFFddVpAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-3499-2071","institution":"Cornell University","correspondingAuthor":true,"prefix":"","firstName":"BAL","middleName":"","lastName":"GOVIND","suffix":""},{"id":569755230,"identity":"f2a219b2-5fc0-4be1-a043-a51f011913aa","order_by":1,"name":"Pablo RAIGOZA","email":"","orcid":"","institution":"Cornell University","correspondingAuthor":false,"prefix":"","firstName":"Pablo","middleName":"","lastName":"RAIGOZA","suffix":""},{"id":569755231,"identity":"82882bed-01a6-4d87-96e5-381263349636","order_by":2,"name":"ALYSSA APSEL","email":"","orcid":"https://orcid.org/0000-0001-9199-2292","institution":"Cornell University","correspondingAuthor":false,"prefix":"","firstName":"ALYSSA","middleName":"","lastName":"APSEL","suffix":""}],"badges":[],"createdAt":"2025-12-14 07:15:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8356393/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8356393/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99854704,"identity":"bd68033f-257c-4a56-8c9e-aa543538966c","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2132270,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/7ac01dd43df92e960903c177.jpg"},{"id":99854706,"identity":"ca4d29cd-7360-4a6b-8351-c4f286ecdc41","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1599127,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/6d27708dd3d4839dcbe32e10.jpg"},{"id":100357552,"identity":"66e220a7-b3f8-4fee-93f1-424fd6d5eac5","added_by":"auto","created_at":"2026-01-16 07:20:02","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2799613,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/db8ac21c7d49092921c2bf66.jpg"},{"id":100056173,"identity":"b62bee25-40d4-45e8-bcbd-841cac770a24","added_by":"auto","created_at":"2026-01-12 13:56:38","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2799613,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/90f5e1b0326e25777bb95214.jpg"},{"id":99854705,"identity":"a0abe79a-0412-4f5e-80aa-b5c8a35b4705","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1510095,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/c2b9fe0d9a6b24838f55897c.jpg"},{"id":99854709,"identity":"1d220e22-d5ec-4067-92fe-53831c875d85","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":4445044,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/69c7755b688a28a704d82272.jpg"},{"id":99854713,"identity":"f084aaa3-fe74-4630-b67c-f1cd409aca9d","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":10376783,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"fig6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/0ece614418acc2f02485aad9.jpg"},{"id":100719007,"identity":"31ea1676-9854-4d40-9bfb-253ffc98d41a","added_by":"auto","created_at":"2026-01-20 19:24:19","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":23227450,"visible":true,"origin":"","legend":"Article File","description":"","filename":"Highspeedfeatureextractionwithintegratedmicrowaveneurons.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1_covered_4dfe6a09-8929-4997-848a-deda38c22104.pdf"},{"id":99854708,"identity":"2af0f136-0b1e-49e1-a0cc-639f7d1fa2dd","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1764883,"visible":true,"origin":"","legend":"Extended Data Fig 1","description":"","filename":"EX1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/7b5aabedfa14bdd9e2d1b6ea.pdf"},{"id":100357813,"identity":"3c5062d4-099f-46a2-a11d-83e69f38d3e1","added_by":"auto","created_at":"2026-01-16 07:20:23","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":6309924,"visible":true,"origin":"","legend":"Extended Data Fig 4","description":"","filename":"EX4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/75735948c8eb74ded093eaee.pdf"},{"id":99854715,"identity":"c0dc821a-c4ff-449f-834f-37813b882cd7","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":11151845,"visible":true,"origin":"","legend":"Extended Data Fig 2","description":"","filename":"EX2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/dc4075b1e3b4225e2eeffe4d.pdf"},{"id":99854712,"identity":"0f1c4f81-0401-4bf9-b8c3-920eba94ce4b","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1440542,"visible":true,"origin":"","legend":"\u003cp\u003eExtended Data Fig 3\u003c/p\u003e","description":"","filename":"EX3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/979189f8ced9a163872eaf10.pdf"},{"id":100056380,"identity":"22b49991-6fe7-4500-8bd5-9ea847103f0a","added_by":"auto","created_at":"2026-01-12 13:58:26","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":6309924,"visible":true,"origin":"","legend":"\u003cp\u003eExtended Data Fig 4\u003c/p\u003e","description":"","filename":"EX4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/f3c9abe90f45572ae2c22fbb.pdf"},{"id":99854711,"identity":"43662880-3dc3-4f87-9208-df885b3ce473","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":7975839,"visible":true,"origin":"","legend":"\u003cp\u003eExtended Data Fig 5\u003c/p\u003e","description":"","filename":"EX5compressed.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/b21e318353de5877acd6bb65.pdf"},{"id":100363980,"identity":"2e71c5be-0b46-48be-a6a5-da9214f514e1","added_by":"auto","created_at":"2026-01-16 07:52:16","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":6309924,"visible":true,"origin":"","legend":"","description":"","filename":"EX4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/d206eda6c021ecc8a872763f.pdf"},{"id":99854714,"identity":"f86b4745-5703-4030-95b5-8e4302d3566e","added_by":"auto","created_at":"2026-01-09 05:18:28","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":19083690,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Information for High-speed feature extraction with integrated microwave neurons\u003c/p\u003e","description":"","filename":"SupplementHighspeedfeatureextraction.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8356393/v1/05a4b9548a2832c8e0636251.pdf"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nB.G. and A.A. have filed a provisional US patent application (number 63/742,208) based on the frontend processor presented in this article. P.R. declares no competing interests.","formattedTitle":"High-speed feature extraction with integrated microwave neurons","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"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":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8356393/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8356393/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Intelligence at high data rates depends on extracting meaning the instant information arrives. However, while real visual, audio and radio signals are fundamentally analog, most computers still fully digitize them before extracting features, which constrains processing speed to the pace of a digital clock and adds latency and power overhead from analog-to-digital conversion.\r\nWe find that, instead, an ensemble of coupled waveguides could extract spectral features of incoming signals instantaneously through interactions between frequency modes across tens of gigahertz. This milliwatt-scale Microwave Neural Network (MNN) uses the incoming signals themselves to reconfigure coupling between waveguides, reshaping its spectrum in real time. By instantly expressing each token’s features across many frequencies, the MNN makes the relationships between successive tokens easier to detect, enabling faster decision-making — without digital preprocessing. Also, when the waveguides are driven directly by gigabit-per-second bitstreams, such as 8-bit pixel data, the MNN produces probabilistic-bits whose bias reflects the input pattern. This acts like an analog form of dithering with lower overhead, where controlled randomness preserves fine detail using fewer bits. As a result, the readout can sample at only one-eigth of the bit rate while still retaining the distinguishing features of the original pattern. This could let satellites downlink richer imagery within tight bandwidth limits and enable radar sensors to classify threats from radio-frequency signals as they arrive—embedding context directly into clockless front-end circuitry.","manuscriptTitle":"High-speed feature extraction with integrated microwave neurons","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-09 05:18:23","doi":"10.21203/rs.3.rs-8356393/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"bf9682c0-c74f-454d-814c-c735d329b19a","owner":[],"postedDate":"January 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":60624226,"name":"Physical sciences/Mathematics and computing/Computer science"},{"id":60624227,"name":"Physical sciences/Engineering/Electrical and electronic engineering"},{"id":60624228,"name":"Physical sciences/Physics/Electronics, photonics and device physics/Electronic and spintronic devices"}],"tags":[],"updatedAt":"2026-03-03T15:41:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-09 05:18:23","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8356393","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8356393","identity":"rs-8356393","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}