{"authors":[{"id":"orcid_______::abfb4b9f9c9e03fca1044d6d5b96e695","fullName":"Kacper Cybinski","name":null,"surname":null,"rank":1,"pid":{"id":{"scheme":"orcid","value":"0000-0003-2600-7473"},"provenance":null}},{"id":"orcid_______::6bacf0e050d82655269541a472722c89","fullName":"Marcin Plodzien","name":null,"surname":null,"rank":2,"pid":{"id":{"scheme":"orcid","value":"0000-0002-0835-1644"},"provenance":null}},{"id":"orcid_______::60cec9973b46c625a6cac41e4131840d","fullName":"Michal Tomza","name":null,"surname":null,"rank":3,"pid":{"id":{"scheme":"orcid","value":"0000-0003-1792-8043"},"provenance":null}},{"id":"orcid_______::a5e7932c42741bfcdebaae104acade96","fullName":"Maciej Lewenstein","name":null,"surname":null,"rank":4,"pid":{"id":{"scheme":"orcid","value":"0000-0002-0210-7800"},"provenance":null}},{"id":"orcid_______::6f72bfab821658ba76dbf4ea005156e7","fullName":"Alexandre Dauphin","name":null,"surname":null,"rank":5,"pid":{"id":{"scheme":"orcid","value":"0000-0003-4996-2561"},"provenance":null}},{"id":"orcid_______::327ddf1576dca872c550bbc6fcf95470","fullName":"Anna Dawid","name":null,"surname":null,"rank":6,"pid":{"id":{"scheme":"orcid","value":"0000-0001-9498-1732"},"provenance":null}}],"openAccessColor":"gold","publiclyFunded":false,"eoscIfGuidelines":null,"type":"publication","language":{"code":"und","label":"Undetermined"},"countries":null,"subjects":[{"subject":{"scheme":"keyword","value":"Computer engineering. Computer hardware"},"provenance":null},{"subject":{"scheme":"keyword","value":"Quantum Physics"},"provenance":null},{"subject":{"scheme":"keyword","value":"out-of-distribution generalization"},"provenance":null},{"subject":{"scheme":"keyword","value":"FOS: Physical sciences"},"provenance":null},{"subject":{"scheme":"FOS","value":"02 engineering and technology"},"provenance":null},{"subject":{"scheme":"keyword","value":"disorder"},"provenance":null},{"subject":{"scheme":"keyword","value":"QA75.5-76.95"},"provenance":null},{"subject":{"scheme":"keyword","value":"Disordered Systems and Neural Networks (cond-mat.dis-nn)"},"provenance":null},{"subject":{"scheme":"keyword","value":"Condensed Matter - Disordered Systems and Neural Networks"},"provenance":null},{"subject":{"scheme":"FOS","value":"01 natural sciences"},"provenance":null},{"subject":{"scheme":"keyword","value":"TK7885-7895"},"provenance":null},{"subject":{"scheme":"keyword","value":"Electronic computers. Computer science"},"provenance":null},{"subject":{"scheme":"FOS","value":"0103 physical sciences"},"provenance":null},{"subject":{"scheme":"keyword","value":"topological phases of matter"},"provenance":null},{"subject":{"scheme":"FOS","value":"0210 nano-technology"},"provenance":null},{"subject":{"scheme":"keyword","value":"interpretability"},"provenance":null},{"subject":{"scheme":"keyword","value":"Quantum Physics (quant-ph)"},"provenance":null}],"mainTitle":"Characterizing out-of-distribution generalization of neural networks: application to the disordered Su–Schrieffer–Heeger model","subTitle":null,"descriptions":["<jats:title>Abstract</jats:title>                <jats:p>Machine learning (ML) is a promising tool for the detection of phases of matter. However, ML models are also known for their black-box construction, which hinders understanding of what they learn from the data and makes their application to novel data risky. Moreover, the central challenge of ML is to ensure its good generalization abilities, i.e. good performance on data outside the training set. Here, we show how the informed use of an interpretability method called class activation mapping, and the analysis of the latent representation of the data with the principal component analysis can increase trust in predictions of a neural network (NN) trained to classify quantum phases. In particular, we show that we can ensure better out-of-distribution (OOD) generalization in the complex classification problem by choosing such an NN that, in the simplified version of the problem, learns a known characteristic of the phase. We also discuss the characteristics of the data representation learned by a network that are predictors of its good OOD generalization. We show this on an example of the topological Su–Schrieffer–Heeger model with and without disorder, which turned out to be surprisingly challenging for NNs trained in a supervised way. This work is an example of how the systematic use of interpretability methods can improve the performance of NNs in scientific problems.</jats:p>"],"publicationDate":"2025-01-22","publisher":"IOP Publishing","embargoEndDate":"2024-06-01","sources":["Crossref","Machine Learning: Science and Technology, Vol 6, Iss 1, p 015014 (2025)"],"formats":null,"contributors":null,"coverages":null,"bestAccessRight":{"code":"c_abf2","label":"OPEN","scheme":"http://vocabularies.coar-repositories.org/documentation/access_rights/"},"container":{"name":"Machine Learning: Science and Technology","issnPrinted":null,"issnOnline":"2632-2153","issnLinking":null,"ep":null,"iss":null,"sp":"015014","vol":"6","edition":null,"conferencePlace":null,"conferenceDate":null},"documentationUrls":null,"codeRepositoryUrl":null,"programmingLanguage":null,"contactPeople":null,"contactGroups":null,"tools":null,"size":null,"version":null,"geoLocations":null,"id":"doi_dedup___::a96bec5ab80770e2d77e6b1e9d0829b8","originalIds":["10.1088/2632-2153/ad9079","50|doiboost____|a96bec5ab80770e2d77e6b1e9d0829b8","50|datacite____::1d280fd478f6559c02e1f724beb1a722","10.48550/arxiv.2406.10012","oai:arXiv.org:2406.10012","50|od________18::68000ccd245a7bef3ae4cdbd74b48ac2","50|dblp________::77d0a26ef55152e1386b9338f7eea198","50|doajarticles::514f1f110b5bef7ba2f09e43270bb860","oai:doaj.org/article:60a2d36ceaa6463c832f109832b87d89"],"pids":[{"scheme":"doi","value":"10.1088/2632-2153/ad9079"},{"scheme":"doi","value":"10.48550/arxiv.2406.10012"},{"scheme":"arXiv","value":"2406.10012"}],"dateOfCollection":null,"lastUpdateTimeStamp":null,"indicators":{"citationImpact":{"citationCount":5.0,"influence":2.4611615E-9,"popularity":5.5309712E-9,"impulse":5.0,"citationClass":"C5","influenceClass":"C5","impulseClass":"C4","popularityClass":"C4"}},"projects":[{"id":"corda_____he::60220ec6b6071552dbe26626539ec9ce","code":"101113690","acronym":"PASQuanS2.1","title":"Programmable Atomic Large-scale Quantum Simulation 2 - SGA1","funder":"European Commission","pids":[{"scheme":"doi","value":"10.3030/101113690"}]},{"id":"corda__h2020::dc1a771ae8ebe96bd047e07e510728a0","code":"847648","acronym":"JUNIOR LEADER","title":"Junior Leader la Caixa Postdoctoral Fellowship Programme: Shaping the new generation of leaders in research","funder":"European Commission","pids":[{"scheme":"doi","value":"10.3030/847648"}]},{"id":"corda__h2020::43979d05c48da9412e5c72f450735d5e","code":"899794","acronym":"OPTOlogic","title":"Optical Topologic Logic","funder":"European Commission","pids":[{"scheme":"doi","value":"10.3030/899794"}]},{"id":"corda__h2020::b890c2abd1dc4d0ca463030a50d3e281","code":"101017733","acronym":"QuantERA II","title":"QuantERA II ERA-NET Cofund in Quantum Technologies","funder":"European Commission","pids":[{"scheme":"doi","value":"10.3030/101017733"}]},{"id":"corda_____he::5982a9992607c81d5b981b167de6e626","code":"101080086","acronym":"NeQST","title":"NExt level Quantum information processing for Science and Technology","funder":"European Commission","pids":[{"scheme":"doi","value":"10.3030/101080086"}]}],"organizations":[{"legalName":"Institute of Photonic Sciences","acronym":"ICFO","id":"openorgs____::6afa298b37a995d4acdff52d6528d053","pids":[{"scheme":"Wikidata","value":"Q3775506"},{"scheme":"wikidata","value":"Q3775506"},{"scheme":"ROR","value":"https://ror.org/03g5ew477"},{"scheme":"OrgReg","value":"ES1035"},{"scheme":"GRID","value":"grid.5853.b"},{"scheme":"fundref","value":"100017170"},{"scheme":"FundRef","value":"100017170"},{"scheme":"OrgRef","value":"47000156"},{"scheme":"ISNI","value":"0000000417571854"},{"scheme":"FundRef","value":"100008050"}],"countries":[{"code":"ES","label":"Spain"}],"websiteurl":"http://www.icfo.eu/"},{"legalName":"University of Warsaw","acronym":"UW","id":"openorgs____::7fe2f66cdc43983c6b24816bfe9cf6a0","pids":[{"scheme":"FundRef","value":"501100006446"},{"scheme":"ROR","value":"https://ror.org/039bjqg32"},{"scheme":"FundRef","value":"501100006445"},{"scheme":"wikidata","value":"Q144488"},{"scheme":"OrgRef","value":"11157563"},{"scheme":"fundref","value":"501100006445"},{"scheme":"GRID","value":"grid.12847.38"},{"scheme":"ISNI","value":"0000000419371290"},{"scheme":"RRID","value":"RRID:SCR_011747"},{"scheme":"PIC","value":"999572294"},{"scheme":"mag_id","value":"4654613"},{"scheme":"Wikidata","value":"Q144488"},{"scheme":"RRID","value":"RRID:nlx_152513"},{"scheme":"fundref","value":"501100006446"}],"countries":[{"code":"PL","label":"Poland"}],"websiteurl":"http://en.uw.edu.pl/"},{"legalName":"PASQAL","acronym":"PASQAL","id":"pending_org_::b147738f93942992f1b1cad9c1be8f88","pids":[{"scheme":"PIC","value":"893652853"}],"countries":[{"code":"FR","label":"France"}],"websiteurl":"http://www.pasqal.com"},{"legalName":"PASQAL SAS","acronym":"PASQAL SAS","id":"pending_org_::ca4f58173319f907aae484485aaeccc7","pids":null,"countries":[{"code":"FR","label":"France"}],"websiteurl":null},{"legalName":"FUNDACIO INSTITUT DE CIENCIES FOTONIQUES","acronym":"ICFO","id":"pending_org_::f8304d66fdd47c56d65cf67bdf44160b","pids":[{"scheme":"PIC","value":"999619436"}],"countries":[{"code":"ES","label":"Spain"}],"websiteurl":"http://www.icfo.eu"},{"legalName":"Fundación Bancaria Caixa d’Estalvis i Pensions de Barcelona","acronym":"Fundación Bancaria Caixa d’Estalvis i Pensions de Barcelona","id":"openorgs____::946641d96b1220ab902ab05429b3d7e7","pids":[{"scheme":"PIC","value":"953757060"},{"scheme":"PIC","value":"998954404"},{"scheme":"ROR","value":"https://ror.org/05r3yzw86"},{"scheme":"fundref","value":"501100019179"},{"scheme":"FundRef","value":"501100019179"},{"scheme":"GRID","value":"grid.434613.7"}],"countries":[{"code":"ES","label":"Spain"}],"websiteurl":"https://www.fundacionbancarialacaixa.org/index_en.html"},{"legalName":"Institució Catalana de Recerca i Estudis Avançats","acronym":"ICREA","id":"openorgs____::75ee0ca9c872b64de338ab062382f810","pids":[{"scheme":"OrgRef","value":"47000151"},{"scheme":"fundref","value":"501100003741"},{"scheme":"Wikidata","value":"Q5051216"},{"scheme":"FundRef","value":"501100003741"},{"scheme":"GRID","value":"grid.425902.8"},{"scheme":"ISNI","value":"000000009601989X"},{"scheme":"ROR","value":"https://ror.org/0371hy230"},{"scheme":"wikidata","value":"Q5051216"},{"scheme":"mag_id","value":"11932220"}],"countries":[{"code":"ES","label":"Spain"}],"websiteurl":"https://www.icrea.cat/en"},{"legalName":"Institución Catalana de Investigación y Estudios Avanzados","acronym":"ICREA","id":"pending_org_::a3c712db42a859f32408277e6f0f2bc9","pids":[{"scheme":"OrgReg","value":"ES4001"}],"countries":[{"code":"ES","label":"Spain"}],"websiteurl":"https://www.icrea.cat/en"},{"legalName":"Flatiron Institute","acronym":"Flatiron Institute","id":"pending_org_::55bafdc85f19c218cc136650ac3bbeca","pids":[{"scheme":"ISNI","value":"0000000474113681"},{"scheme":"wikidata","value":"Q30594395"},{"scheme":"Wikidata","value":"Q30594395"},{"scheme":"ROR","value":"https://ror.org/00sekdz59"}],"countries":[{"code":"US","label":"United States"}],"websiteurl":"https://www.simonsfoundation.org/flatiron"},{"legalName":"University of Warsaw, Faculty of Physics","acronym":"University of Warsaw, Faculty of Physics","id":"pending_org_::a745cdc5516fc3be309d9f15020500f5","pids":null,"countries":[{"code":"PL","label":"Poland"}],"websiteurl":null},{"legalName":"Barcelona Institute for Science and Technology","acronym":"BIST","id":"openorgs____::b7949bd169e349b917447bb919b18f74","pids":[{"scheme":"FundRef","value":"100017168"},{"scheme":"ROR","value":"https://ror.org/03kpps236"},{"scheme":"Wikidata","value":"Q30267112"},{"scheme":"fundref","value":"100017168"},{"scheme":"GRID","value":"grid.473715.3"},{"scheme":"ISNI","value":"0000000464757299"},{"scheme":"wikidata","value":"Q30267112"}],"countries":[{"code":"ES","label":"Spain"}],"websiteurl":"http://bist.eu/"}],"communities":null,"collectedFrom":[{"key":"openaire____::9e3be59865b2c1c335d32dae2fe7b254","value":"Datacite"},{"key":"driver______::bee53aa31dc2cbb538c10c2b65fa5824","value":"DOAJ"},{"key":"openaire____::081b82f96300b6a6e3d282bad31cb6e2","value":"Crossref"},{"key":"openaire____::d8b68cc0a53121f6f896883a7d60c1db","value":"DBLP"},{"key":"opendoar____::6f4922f45568161a8cdf4ad2299f6d23","value":"arXiv.org e-Print Archive"}],"instances":[{"pids":[{"scheme":"doi","value":"10.1088/2632-2153/ad9079"}],"license":"CC BY","accessRight":{"code":"c_abf2","label":"OPEN","scheme":"http://vocabularies.coar-repositories.org/documentation/access_rights/","openAccessRoute":"gold"},"type":"Article","urls":["https://doi.org/10.1088/2632-2153/ad9079"],"publicationDate":"2025-01-22","refereed":"peerReviewed","hostedBy":{"key":"doajarticles::4875bf86e6e2dc8d9f0823a76cbf98f6","value":"Machine Learning: Science and Technology"},"collectedFrom":{"key":"openaire____::081b82f96300b6a6e3d282bad31cb6e2","value":"Crossref"}},{"pids":[{"scheme":"doi","value":"10.48550/arxiv.2406.10012"}],"license":"CC BY","type":"Article","urls":["https://dx.doi.org/10.48550/arxiv.2406.10012"],"publicationDate":"2024-01-01","refereed":"nonPeerReviewed","hostedBy":{"key":"openaire____::55045bd2a65019fd8e6741a755395c8c","value":"Unknown Repository"},"collectedFrom":{"key":"openaire____::9e3be59865b2c1c335d32dae2fe7b254","value":"Datacite"}},{"pids":[{"scheme":"arXiv","value":"2406.10012"}],"alternateIdentifiers":[{"scheme":"doi","value":"10.1088/2632-2153/ad9079"}],"accessRight":{"code":"c_abf2","label":"OPEN","scheme":"http://vocabularies.coar-repositories.org/documentation/access_rights/","openAccessRoute":null},"type":"Preprint","urls":["http://arxiv.org/abs/2406.10012"],"publicationDate":"2024-06-14","refereed":"nonPeerReviewed","hostedBy":{"key":"opendoar____::6f4922f45568161a8cdf4ad2299f6d23","value":"arXiv.org e-Print Archive"},"collectedFrom":{"key":"opendoar____::6f4922f45568161a8cdf4ad2299f6d23","value":"arXiv.org e-Print Archive"}},{"alternateIdentifiers":[{"scheme":"doi","value":"10.1088/2632-2153/ad9079"}],"type":"Article","urls":["https://dblp.org/rec/journals/mlst/CybinskiPTLDD25.html","https://doi.org/10.1088/2632-2153/ad9079"],"publicationDate":"2025-01-01","refereed":"nonPeerReviewed","hostedBy":{"key":"openaire____::d8b68cc0a53121f6f896883a7d60c1db","value":"DBLP"},"collectedFrom":{"key":"openaire____::d8b68cc0a53121f6f896883a7d60c1db","value":"DBLP"}},{"alternateIdentifiers":[{"scheme":"doi","value":"10.1088/2632-2153/ad9079"}],"accessRight":{"code":"c_abf2","label":"OPEN","scheme":"http://vocabularies.coar-repositories.org/documentation/access_rights/","openAccessRoute":"gold"},"type":"Article","urls":["https://doaj.org/article/60a2d36ceaa6463c832f109832b87d89"],"publicationDate":"2025-01-01","refereed":"nonPeerReviewed","hostedBy":{"key":"doajarticles::4875bf86e6e2dc8d9f0823a76cbf98f6","value":"Machine Learning: Science and Technology"},"collectedFrom":{"key":"driver______::bee53aa31dc2cbb538c10c2b65fa5824","value":"DOAJ"}}],"links":[{"header":{"relationType":"resultProject","relationClass":"isProducedBy","relatedIdentifier":"corda_____he::60220ec6b6071552dbe26626539ec9ce","relatedRecordType":"project","relationProvenance":"iis","trust":"0.897"},"collectedfrom":[{"dsId":"openaire____::3f264f93cf3b0cfc4ede188a6300455c","dsName":"CORDA - COmmon Research DAta Warehouse - Horizon Europe"}],"projectTitle":"Programmable Atomic Large-scale Quantum Simulation 2 - SGA1","code":"101113690","funding":{"funder":{"id":"ec__________::EC","shortname":"EC","name":"European Commission","jurisdiction":{"code":"EU","label":"European Union"},"pid":null},"level0":{"id":"ec__________::EC::HE","description":"Horizon Europe Framework Programme","name":"HE"},"level1":{"id":"ec__________::EC::HE::HORIZON-RIA","description":"HORIZON  Research and Innovation Actions","name":"HORIZON-RIA"},"level2":{"id":null,"description":null,"name":null}},"startDate":"2023-04-01","endDate":"2026-09-30"},{"header":{"relationType":"resultProject","relationClass":"isProducedBy","relatedIdentifier":"corda__h2020::dc1a771ae8ebe96bd047e07e510728a0","relatedRecordType":"project","relationProvenance":"iis","trust":"0.897"},"collectedfrom":[{"dsId":"openaire____::a55eb91348674d853191f4f4fd73d078","dsName":"CORDA - COmmon Research DAta Warehouse - Horizon 2020"}],"projectTitle":"Junior Leader la Caixa Postdoctoral Fellowship Programme: Shaping the new generation of leaders in research","code":"847648","funding":{"funder":{"id":"ec__________::EC","shortname":"EC","name":"European Commission","jurisdiction":{"code":"EU","label":"European Union"},"pid":null},"level0":{"id":"ec__________::EC::H2020","description":"Horizon 2020 Framework Programme","name":"H2020"},"level1":{"id":"ec__________::EC::H2020::MSCA-COFUND-FP","description":"Fellowship programmes","name":"MSCA-COFUND-FP"},"level2":{"id":null,"description":null,"name":null}},"startDate":"2019-07-08","endDate":"2025-02-07"},{"header":{"relationType":"resultProject","relationClass":"isProducedBy","relatedIdentifier":"corda__h2020::43979d05c48da9412e5c72f450735d5e","relatedRecordType":"project","relationProvenance":"iis","trust":"0.897"},"collectedfrom":[{"dsId":"openaire____::a55eb91348674d853191f4f4fd73d078","dsName":"CORDA - COmmon Research DAta Warehouse - Horizon 2020"}],"projectTitle":"Optical Topologic Logic","code":"899794","funding":{"funder":{"id":"ec__________::EC","shortname":"EC","name":"European Commission","jurisdiction":{"code":"EU","label":"European Union"},"pid":null},"level0":{"id":"ec__________::EC::H2020","description":"Horizon 2020 Framework Programme","name":"H2020"},"level1":{"id":"ec__________::EC::H2020::RIA","description":"Research and Innovation action","name":"RIA"},"level2":{"id":null,"description":null,"name":null}},"startDate":"2020-09-01","endDate":"2025-08-31"},{"header":{"relationType":"resultProject","relationClass":"isProducedBy","relatedIdentifier":"corda__h2020::b890c2abd1dc4d0ca463030a50d3e281","relatedRecordType":"project","relationProvenance":"iis","trust":"0.897"},"collectedfrom":[{"dsId":"openaire____::a55eb91348674d853191f4f4fd73d078","dsName":"CORDA - COmmon Research DAta Warehouse - Horizon 2020"}],"projectTitle":"QuantERA II ERA-NET Cofund in Quantum Technologies","code":"101017733","funding":{"funder":{"id":"ec__________::EC","shortname":"EC","name":"European Commission","jurisdiction":{"code":"EU","label":"European Union"},"pid":null},"level0":{"id":"ec__________::EC::H2020","description":"Horizon 2020 Framework Programme","name":"H2020"},"level1":{"id":"ec__________::EC::H2020::ERA-NET-Cofund","description":"ERA-NET Cofund","name":"ERA-NET-Cofund"},"level2":{"id":null,"description":null,"name":null}},"startDate":"2021-01-01","endDate":"2026-12-31"},{"header":{"relationType":"resultProject","relationClass":"isProducedBy","relatedIdentifier":"corda_____he::5982a9992607c81d5b981b167de6e626","relatedRecordType":"project","relationProvenance":"iis","trust":"0.897"},"collectedfrom":[{"dsId":"openaire____::3f264f93cf3b0cfc4ede188a6300455c","dsName":"CORDA - COmmon Research DAta Warehouse - Horizon Europe"}],"projectTitle":"NExt level Quantum information processing for Science and Technology","code":"101080086","funding":{"funder":{"id":"ec__________::EC","shortname":"EC","name":"European Commission","jurisdiction":{"code":"EU","label":"European Union"},"pid":null},"level0":{"id":"ec__________::EC::HE","description":"Horizon Europe Framework Programme","name":"HE"},"level1":{"id":"ec__________::EC::HE::HORIZON-RIA","description":"HORIZON  Research and Innovation Actions","name":"HORIZON-RIA"},"level2":{"id":null,"description":null,"name":null}},"startDate":"2022-11-01","endDate":"2026-02-28"}],"otherTitles":null,"green":true,"inDiamondJournal":false,"isGreen":true,"isInDiamondJournal":false}