Evidential Deep Learning for Interatomic Potentials

Evidential Deep Learning for Interatomic Potentials | 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 Evidential Deep Learning for Interatomic Potentials Mao Su, Han Xu, Taoyong Cui, Chenyu Tang, Dongzhan Zhou, Yuqiang Li, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4793228/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 Machine learning interatomic potentials (MLIPs) have been widely used to facilitate large scale molecular simulations with ab initio level accuracy. However, MLIP-based molecular simulations frequently encounter the issue of collapse due to decreased prediction accuracy for out-of-distribution (OOD) data. To mitigate this issue, it is crucial to enrich the training set with active learning, where uncertainty estimation serves as an effective method for identifying and collecting OOD data. Therefore, a feasible method for uncertainty estimation in MLIPs is desired. The existing methods either require expensive computations or compromise prediction accuracy. In this work, we introduce evidential deep learning for interatomic potentials (eIP) with a physics-inspired design. Our experiments demonstrate that eIP consistently generates reliable uncertainties without incurring notable additional computational costs, while the prediction accuracy remains unchanged. Furthermore, we present an eIP-based active learning workflow, where eIP is used not only to estimate the uncertainty of molecular data but also to perform uncertainty-driven dynamics simulations. Our findings show that eIP enables efficient sampling for a more diverse dataset, thereby advancing the feasibility of MLIP-based molecular simulations. Physical sciences/Materials science/Theory and computation/Atomistic models Physical sciences/Mathematics and computing/Computational science Physical sciences/Chemistry/Theoretical chemistry/Molecular dynamics Full Text Additional Declarations There is NO Competing Interest. Supplementary Files eIPSI.pdf 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-4793228","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":343035941,"identity":"16e01289-3125-46ec-a6f7-93df19290a48","order_by":0,"name":"Mao Su","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYNCCCgZmPgYekrScYWBmI00LYxsDA/Fa5GdkJ374OO8OOxv72QMMH/fUMvDPbsCvxeBG7mbJmdueMbPx5CUwznh2nEHizgECWiRyN0jzbjsM9EuOATPPgWNAkQRCDsvd/Jt3DlAL/xsitTDcyN0mzdsA1CIBtqWGsBaDM2+3Wc44BvSLxLuEgzMOHOCRuEHIYe25m298qLmTzM+fe/DBhwN1cvwzCDkMAg4kg0kGhsNER+gBOyijjlgdo2AUjIJRMIIAAKbCQplFp95iAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-2066-881X","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":true,"prefix":"","firstName":"Mao","middleName":"","lastName":"Su","suffix":""},{"id":343035942,"identity":"d27f23ba-d4f4-4221-9711-e7cc94019f23","order_by":1,"name":"Han Xu","email":"","orcid":"https://orcid.org/0000-0002-2751-9668","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Han","middleName":"","lastName":"Xu","suffix":""},{"id":343035943,"identity":"4e1ee739-d007-4bba-bc8e-884a0e3a7625","order_by":2,"name":"Taoyong Cui","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Taoyong","middleName":"","lastName":"Cui","suffix":""},{"id":343035944,"identity":"b03e733f-f3b6-435b-b793-9deab38e86f2","order_by":3,"name":"Chenyu Tang","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Chenyu","middleName":"","lastName":"Tang","suffix":""},{"id":343035945,"identity":"f8a36f14-4917-4e8c-a118-8e74c61638d4","order_by":4,"name":"Dongzhan Zhou","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Dongzhan","middleName":"","lastName":"Zhou","suffix":""},{"id":343035946,"identity":"d4793561-3874-4dbb-8b05-ddc88935b11c","order_by":5,"name":"Yuqiang Li","email":"","orcid":"https://orcid.org/0000-0001-6756-6154","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Yuqiang","middleName":"","lastName":"Li","suffix":""},{"id":343035947,"identity":"83649a0d-a53f-4a67-ba7d-3cbc5e79db91","order_by":6,"name":"Xiang Gao","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Xiang","middleName":"","lastName":"Gao","suffix":""},{"id":343035948,"identity":"c153b33f-1fb4-40b1-8114-d4026c8cd5b5","order_by":7,"name":"Xin-Gao Gong","email":"","orcid":"","institution":"Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xin-Gao","middleName":"","lastName":"Gong","suffix":""},{"id":343035949,"identity":"f2ad309f-ffa1-486f-9e5a-e5fee5f35a36","order_by":8,"name":"Wanli Ouyang","email":"","orcid":"","institution":"Shanghai AI lab","correspondingAuthor":false,"prefix":"","firstName":"Wanli","middleName":"","lastName":"Ouyang","suffix":""},{"id":343035950,"identity":"8d7afec6-f4a5-48b3-a043-2d01a48d2827","order_by":9,"name":"Shufei Zhang","email":"","orcid":"","institution":"Shanghai Artificial Intelligence Laboratory","correspondingAuthor":false,"prefix":"","firstName":"Shufei","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2024-07-24 07:31:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4793228/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4793228/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":65079319,"identity":"5229051f-e125-4005-be37-f87a91215cf9","added_by":"auto","created_at":"2024-09-23 11:40:27","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5916407,"visible":true,"origin":"","legend":"","description":"","filename":"eIPmain.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4793228/v1_covered_26bab1f7-5b7f-4039-8d2c-7fd97f4fed6c.pdf"},{"id":62995375,"identity":"cab2f913-cbc6-47fc-b9db-2f63b073073f","added_by":"auto","created_at":"2024-08-22 01:12:00","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":13349463,"visible":true,"origin":"","legend":"","description":"","filename":"eIPSI.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4793228/v1/e3db2ad77f2e615d911d53d1.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Evidential Deep Learning for Interatomic Potentials","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":"","lastPublishedDoi":"10.21203/rs.3.rs-4793228/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4793228/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Machine learning interatomic potentials (MLIPs) have been widely used to facilitate large scale molecular simulations with ab initio level accuracy. However, MLIP-based molecular simulations frequently encounter the issue of collapse due to decreased prediction accuracy for out-of-distribution (OOD) data. To mitigate this issue, it is crucial to enrich the training set with active learning, where uncertainty estimation serves as an effective method for identifying and collecting OOD data. Therefore, a feasible method for uncertainty estimation in MLIPs is desired. The existing methods either require expensive computations or compromise prediction accuracy. In this work, we introduce evidential deep learning for interatomic potentials (eIP) with a physics-inspired design. Our experiments demonstrate that eIP consistently generates reliable uncertainties without incurring notable additional computational costs, while the prediction accuracy remains unchanged. Furthermore, we present an eIP-based active learning workflow, where eIP is used not only to estimate the uncertainty of molecular data but also to perform uncertainty-driven dynamics simulations. Our findings show that eIP enables efficient sampling for a more diverse dataset, thereby advancing the feasibility of MLIP-based molecular simulations.","manuscriptTitle":"Evidential Deep Learning for Interatomic Potentials","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-08-22 01:11:55","doi":"10.21203/rs.3.rs-4793228/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":"4647730c-b06e-4cd2-a456-bb976b5f0b94","owner":[],"postedDate":"August 22nd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":36344175,"name":"Physical sciences/Materials science/Theory and computation/Atomistic models"},{"id":36344176,"name":"Physical sciences/Mathematics and computing/Computational science"},{"id":36344177,"name":"Physical sciences/Chemistry/Theoretical chemistry/Molecular dynamics"}],"tags":[],"updatedAt":"2025-05-23T16:50:15+00:00","versionOfRecord":[],"versionCreatedAt":"2024-08-22 01:11:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4793228","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4793228","identity":"rs-4793228","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}