Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks

Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile beha...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2021-03
Hauptverfasser: Schwalbe-Koda, Daniel, Aik Rui Tan, Gómez-Bombarelli, Rafael
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Schwalbe-Koda, Daniel
Aik Rui Tan
Gómez-Bombarelli, Rafael
description Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile behavior when extrapolating. Uncertainty quantification approaches can flag atomic configurations for which prediction confidence is low, but arriving at such uncertain regions requires expensive sampling of the NN phase space, often using atomistic simulations. Here, we exploit automatic differentiation to drive atomistic systems towards high-likelihood, high-uncertainty configurations without the need for molecular dynamics simulations. By performing adversarial attacks on an uncertainty metric, informative geometries that expand the training domain of NNs are sampled. When combined to an active learning loop, this approach bootstraps and improves NN potentials while decreasing the number of calls to the ground truth method. This efficiency is demonstrated on sampling of kinetic barriers and collective variables in molecules, and can be extended to any NN potential architecture and materials system.
doi_str_mv 10.48550/arxiv.2101.11588
format Article
fullrecord <record><control><sourceid>proquest_arxiv</sourceid><recordid>TN_cdi_arxiv_primary_2101_11588</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2482385497</sourcerecordid><originalsourceid>FETCH-LOGICAL-a527-9d1a8bf8a43866736dda22185b261cecd339ac012a7733eb45eadfd9cd56f44b3</originalsourceid><addsrcrecordid>eNotkMtOwzAURC0kJKrSD2CFJdYpfibOEpVHkSqx6YpNdGPfFJc8iu0U-veUltVsjkYzh5AbzubKaM3uIfz4_Vxwxueca2MuyERIyTOjhLgisxi3jDGRF0JrOSHvj75pMGCfPNQt0gjdrvX9hg4N7YYW7dhCoBscOkzBY6TfPn3QsbcYEvg-HbIaIjoKbo8hQvDQUkgJ7Ge8JpcNtBFn_zkl6-en9WKZrd5eXhcPqwy0KLLScTB1Y0BJk-eFzJ0DIbjRtci5ReukLMEyLqAopMRaaQTXuNI6nTdK1XJKbs-1p-PVLvgOwqH6E1CdBByJuzOxC8PXiDFV22EM_XFTJZQR0mhVFvIX4rJgJg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2482385497</pqid></control><display><type>article</type><title>Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks</title><source>arXiv.org</source><source>Free E- Journals</source><creator>Schwalbe-Koda, Daniel ; Aik Rui Tan ; Gómez-Bombarelli, Rafael</creator><creatorcontrib>Schwalbe-Koda, Daniel ; Aik Rui Tan ; Gómez-Bombarelli, Rafael</creatorcontrib><description>Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile behavior when extrapolating. Uncertainty quantification approaches can flag atomic configurations for which prediction confidence is low, but arriving at such uncertain regions requires expensive sampling of the NN phase space, often using atomistic simulations. Here, we exploit automatic differentiation to drive atomistic systems towards high-likelihood, high-uncertainty configurations without the need for molecular dynamics simulations. By performing adversarial attacks on an uncertainty metric, informative geometries that expand the training domain of NNs are sampled. When combined to an active learning loop, this approach bootstraps and improves NN potentials while decreasing the number of calls to the ground truth method. This efficiency is demonstrated on sampling of kinetic barriers and collective variables in molecules, and can be extended to any NN potential architecture and materials system.</description><identifier>EISSN: 2331-8422</identifier><identifier>DOI: 10.48550/arxiv.2101.11588</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Active learning ; Computer Science - Learning ; Domains ; Electronic structure ; Learning ; Neural networks ; Physics - Chemical Physics ; Physics - Statistical Mechanics ; Potential energy ; Training ; Uncertainty</subject><ispartof>arXiv.org, 2021-03</ispartof><rights>2021. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,780,881,27904</link.rule.ids><backlink>$$Uhttps://doi.org/10.1038/s41467-021-25342-8$$DView published paper (Access to full text may be restricted)$$Hfree_for_read</backlink><backlink>$$Uhttps://doi.org/10.48550/arXiv.2101.11588$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Schwalbe-Koda, Daniel</creatorcontrib><creatorcontrib>Aik Rui Tan</creatorcontrib><creatorcontrib>Gómez-Bombarelli, Rafael</creatorcontrib><title>Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks</title><title>arXiv.org</title><description>Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile behavior when extrapolating. Uncertainty quantification approaches can flag atomic configurations for which prediction confidence is low, but arriving at such uncertain regions requires expensive sampling of the NN phase space, often using atomistic simulations. Here, we exploit automatic differentiation to drive atomistic systems towards high-likelihood, high-uncertainty configurations without the need for molecular dynamics simulations. By performing adversarial attacks on an uncertainty metric, informative geometries that expand the training domain of NNs are sampled. When combined to an active learning loop, this approach bootstraps and improves NN potentials while decreasing the number of calls to the ground truth method. This efficiency is demonstrated on sampling of kinetic barriers and collective variables in molecules, and can be extended to any NN potential architecture and materials system.</description><subject>Active learning</subject><subject>Computer Science - Learning</subject><subject>Domains</subject><subject>Electronic structure</subject><subject>Learning</subject><subject>Neural networks</subject><subject>Physics - Chemical Physics</subject><subject>Physics - Statistical Mechanics</subject><subject>Potential energy</subject><subject>Training</subject><subject>Uncertainty</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><sourceid>GOX</sourceid><recordid>eNotkMtOwzAURC0kJKrSD2CFJdYpfibOEpVHkSqx6YpNdGPfFJc8iu0U-veUltVsjkYzh5AbzubKaM3uIfz4_Vxwxueca2MuyERIyTOjhLgisxi3jDGRF0JrOSHvj75pMGCfPNQt0gjdrvX9hg4N7YYW7dhCoBscOkzBY6TfPn3QsbcYEvg-HbIaIjoKbo8hQvDQUkgJ7Ge8JpcNtBFn_zkl6-en9WKZrd5eXhcPqwy0KLLScTB1Y0BJk-eFzJ0DIbjRtci5ReukLMEyLqAopMRaaQTXuNI6nTdK1XJKbs-1p-PVLvgOwqH6E1CdBByJuzOxC8PXiDFV22EM_XFTJZQR0mhVFvIX4rJgJg</recordid><startdate>20210329</startdate><enddate>20210329</enddate><creator>Schwalbe-Koda, Daniel</creator><creator>Aik Rui Tan</creator><creator>Gómez-Bombarelli, Rafael</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20210329</creationdate><title>Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks</title><author>Schwalbe-Koda, Daniel ; Aik Rui Tan ; Gómez-Bombarelli, Rafael</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a527-9d1a8bf8a43866736dda22185b261cecd339ac012a7733eb45eadfd9cd56f44b3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Active learning</topic><topic>Computer Science - Learning</topic><topic>Domains</topic><topic>Electronic structure</topic><topic>Learning</topic><topic>Neural networks</topic><topic>Physics - Chemical Physics</topic><topic>Physics - Statistical Mechanics</topic><topic>Potential energy</topic><topic>Training</topic><topic>Uncertainty</topic><toplevel>online_resources</toplevel><creatorcontrib>Schwalbe-Koda, Daniel</creatorcontrib><creatorcontrib>Aik Rui Tan</creatorcontrib><creatorcontrib>Gómez-Bombarelli, Rafael</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection><collection>arXiv Computer Science</collection><collection>arXiv.org</collection><jtitle>arXiv.org</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Schwalbe-Koda, Daniel</au><au>Aik Rui Tan</au><au>Gómez-Bombarelli, Rafael</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks</atitle><jtitle>arXiv.org</jtitle><date>2021-03-29</date><risdate>2021</risdate><eissn>2331-8422</eissn><abstract>Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile behavior when extrapolating. Uncertainty quantification approaches can flag atomic configurations for which prediction confidence is low, but arriving at such uncertain regions requires expensive sampling of the NN phase space, often using atomistic simulations. Here, we exploit automatic differentiation to drive atomistic systems towards high-likelihood, high-uncertainty configurations without the need for molecular dynamics simulations. By performing adversarial attacks on an uncertainty metric, informative geometries that expand the training domain of NNs are sampled. When combined to an active learning loop, this approach bootstraps and improves NN potentials while decreasing the number of calls to the ground truth method. This efficiency is demonstrated on sampling of kinetic barriers and collective variables in molecules, and can be extended to any NN potential architecture and materials system.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><doi>10.48550/arxiv.2101.11588</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2021-03
issn 2331-8422
language eng
recordid cdi_arxiv_primary_2101_11588
source arXiv.org; Free E- Journals
subjects Active learning
Computer Science - Learning
Domains
Electronic structure
Learning
Neural networks
Physics - Chemical Physics
Physics - Statistical Mechanics
Potential energy
Training
Uncertainty
title Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-27T05%3A25%3A17IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_arxiv&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Differentiable%20sampling%20of%20molecular%20geometries%20with%20uncertainty-based%20adversarial%20attacks&rft.jtitle=arXiv.org&rft.au=Schwalbe-Koda,%20Daniel&rft.date=2021-03-29&rft.eissn=2331-8422&rft_id=info:doi/10.48550/arxiv.2101.11588&rft_dat=%3Cproquest_arxiv%3E2482385497%3C/proquest_arxiv%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2482385497&rft_id=info:pmid/&rfr_iscdi=true