{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,2,21]],"date-time":"2025-02-21T23:58:43Z","timestamp":1740182323943,"version":"3.37.3"},"reference-count":51,"publisher":"IOP Publishing","issue":"4","license":[{"start":{"date-parts":[[2020,9,11]],"date-time":"2020-09-11T00:00:00Z","timestamp":1599782400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2020,9,11]],"date-time":"2020-09-11T00:00:00Z","timestamp":1599782400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"funder":[{"DOI":"10.13039\/100011039","name":"Intelligence Advanced Research Projects Activity","doi-asserted-by":"crossref","id":[{"id":"10.13039\/100011039","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2020,12,1]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>Known for their ability to identify hidden patterns in data, artificial neural networks are among the most powerful machine learning tools. Most notably, neural networks have played a central role in identifying states of matter and phase transitions across condensed matter physics. To date, most studies have focused on systems where different phases of matter and their phase transitions are known, and thus the performance of neural networks is well controlled. While neural networks present an exciting new tool to detect new phases of matter, here we demonstrate that when the training sets are poisoned (i.e. poor training data or mislabeled data) it is easy for neural networks to make misleading predictions.<\/jats:p>","DOI":"10.1088\/2632-2153\/aba821","type":"journal-article","created":{"date-parts":[[2020,7,21]],"date-time":"2020-07-21T22:45:52Z","timestamp":1595371552000},"page":"045001","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Machine learning in physics: the pitfalls of poisoned training sets"],"prefix":"10.1088","volume":"1","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6626-6587","authenticated-orcid":false,"given":"Chao","family":"Fang","sequence":"first","affiliation":[]},{"given":"Amin","family":"Barzeger","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3341-9943","authenticated-orcid":false,"given":"Helmut G","family":"Katzgraber","sequence":"additional","affiliation":[]}],"member":"266","published-online":{"date-parts":[[2020,9,11]]},"reference":[{"year":"2009","author":"Haykin","key":"mlstaba821bib1"},{"year":"2016","author":"Goodfellow","key":"mlstaba821bib2"},{"year":"2006","author":"Bishop","key":"mlstaba821bib3"},{"key":"mlstaba821bib4","doi-asserted-by":"publisher","first-page":"105","DOI":"10.1140\/epje\/i2011-11105-9","volume":"34","author":"Ronhovde","year":"2011","journal-title":"The European Physical Journal E"},{"first-page":"p 115","year":"2016","author":"Nussinov","key":"mlstaba821bib5"},{"key":"mlstaba821bib6","doi-asserted-by":"publisher","first-page":"431","DOI":"10.1038\/nphys4035","volume":"13","author":"Carrasquilla","year":"2017","journal-title":"Nat. Phys."},{"key":"mlstaba821bib7","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevX.7.031038","volume":"7","author":"Ch\u2019ng","year":"2017","journal-title":"Phys. Rev. X"},{"key":"mlstaba821bib8","doi-asserted-by":"publisher","DOI":"10.7566\/JPSJ.86.063001","volume":"86","author":"Tanaka","year":"2017","journal-title":"J. Phys. Soc. Jpn"},{"key":"mlstaba821bib9","doi-asserted-by":"publisher","first-page":"083A04","DOI":"10.1093\/ptep\/ptz082","volume":"2019","author":"Kashiwa","year":"2019","journal-title":"Prog. Theor. Exp. Phys."},{"key":"mlstaba821bib10","doi-asserted-by":"publisher","first-page":"1307","DOI":"10.1137\/0907087","volume":"7","author":"Santosa","year":"1986","journal-title":"SIAM J. Sci. Stat. Comput."},{"key":"mlstaba821bib11","doi-asserted-by":"crossref","first-page":"267","DOI":"10.1111\/j.2517-6161.1996.tb02080.x","volume":"58","author":"Tibshirani","year":"1996","journal-title":"J. Royal Stat. Soc. B"},{"key":"mlstaba821bib12","doi-asserted-by":"publisher","first-page":"5262","DOI":"10.1109\/TSP.2010.2055862","volume":"58","author":"Mateos","year":"2010","journal-title":"Trans. Sig. Proc."},{"key":"mlstaba821bib13","first-page":"1935","volume":"6","author":"Quinonero Candela","year":"2005","journal-title":"Journal of Machine Learning Research"},{"edition":"2nd ed","year":"2014","author":"Rokach","key":"mlstaba821bib14"},{"year":"2014","author":"Shalev-Shwartz","key":"mlstaba821bib15"},{"key":"mlstaba821bib16","doi-asserted-by":"publisher","first-page":"609","DOI":"10.1016\/S0304-3975(01)00011-1","volume":"270","author":"Mehta","year":"2002","journal-title":"Theor. Comput. Sci."},{"year":"2013","author":"James","key":"mlstaba821bib17"},{"journal-title":"A practical guide to support vector classification (Taipei: National Taiwan University)","year":"2010","author":"Hsu","key":"mlstaba821bib18"},{"first-page":"p 185","year":"1999","author":"Platt","key":"mlstaba821bib19"},{"key":"mlstaba821bib20","doi-asserted-by":"publisher","first-page":"2560","DOI":"10.1016\/j.ymssp.2006.12.007","volume":"21","author":"Widodo","year":"2007","journal-title":"Mech. Syst. Signal Process."},{"first-page":"p 137","year":"1998","author":"Joachims","key":"mlstaba821bib21"},{"first-page":"p 255","year":"1998","author":"LeCun","key":"mlstaba821bib22"},{"key":"mlstaba821bib23","doi-asserted-by":"publisher","first-page":"4790","DOI":"10.1364\/AO.29.004790","volume":"29","author":"Zhang","year":"1990","journal-title":"Appl. Opt."},{"key":"mlstaba821bib24","doi-asserted-by":"publisher","DOI":"10.1088\/1742-5468\/ab9e60","volume":"2020","author":"Munoz-Bauza","year":"2020","journal-title":"J. Stat. Mech."},{"key":"mlstaba821bib25","doi-asserted-by":"publisher","DOI":"10.1088\/1742-5468\/2010\/06\/P06026","volume":"2010","author":"Alvarez Ba\u00f1os","year":"2010","journal-title":"J. Stat. Mech."},{"key":"mlstaba821bib26","doi-asserted-by":"publisher","first-page":"6452","DOI":"10.1073\/pnas.1203295109","volume":"109","author":"Ba\u00f1os","year":"2012","journal-title":"Proc. Natl. Acad. Sci. U.S.A."},{"key":"mlstaba821bib27","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevE.89.032140","volume":"89","author":"Baity-Jesi","year":"2014","journal-title":"Phys. Rev. E"},{"year":"2017","author":"Steinhardt","key":"mlstaba821bib28"},{"year":"2018","author":"Jagielski","key":"mlstaba821bib29"},{"key":"mlstaba821bib30","first-page":"1452","author":"Alfeld","year":"2016","journal-title":"AAAI"},{"year":"2019","author":"Shi","key":"mlstaba821bib31"},{"article-title":"Vulnerability of machine learning phases of matter","year":"2019","author":"Jiang","key":"mlstaba821bib32"},{"year":"2008","author":"Nelson","key":"mlstaba821bib33"},{"year":"2014","author":"Newell","key":"mlstaba821bib34"},{"key":"mlstaba821bib35","doi-asserted-by":"publisher","first-page":"965","DOI":"10.1088\/0305-4608\/5\/5\/017","volume":"5","author":"Edwards","year":"1975","journal-title":"J. Phys. F.: Met. Phys."},{"key":"mlstaba821bib36","doi-asserted-by":"publisher","first-page":"801","DOI":"10.1103\/RevModPhys.58.801","volume":"58","author":"Binder","year":"1986","journal-title":"Rev. Mod. Phys."},{"year":"1987","author":"M\u00e9zard","key":"mlstaba821bib37"},{"year":"1998","author":"Young","key":"mlstaba821bib38"},{"year":"2013","author":"Stein","key":"mlstaba821bib39"},{"journal-title":"TensorFlow: A System for Large-Scale Machine Learning","year":"2016","author":"Abadi","key":"mlstaba821bib40"},{"key":"mlstaba821bib41","doi-asserted-by":"publisher","first-page":"1792","DOI":"10.1103\/PhysRevLett.35.1792","volume":"35","author":"Sherrington","year":"1975","journal-title":"Phys. Rev. Lett."},{"key":"mlstaba821bib42","doi-asserted-by":"publisher","first-page":"1101","DOI":"10.1088\/0305-4470\/13\/3\/042","volume":"13","author":"Parisi","year":"1980","journal-title":"J. Phys. A"},{"key":"mlstaba821bib43","doi-asserted-by":"publisher","first-page":"1946","DOI":"10.1103\/PhysRevLett.50.1946","volume":"50","author":"Parisi","year":"1983","journal-title":"Phys. Rev. Lett."},{"key":"mlstaba821bib44","doi-asserted-by":"publisher","first-page":"1604","DOI":"10.1143\/JPSJ.65.1604","volume":"65","author":"Hukushima","year":"1996","journal-title":"J. Phys. Soc. Jpn."},{"key":"mlstaba821bib45","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevB.73.224432","volume":"73","author":"Katzgraber","year":"2006","journal-title":"Phys. Rev. B"},{"key":"mlstaba821bib46","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevB.58.14852","volume":"58","author":"Marinari","year":"1998","journal-title":"Phys. Rev. B"},{"key":"mlstaba821bib47","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevB.72.014462","volume":"72","author":"Katzgraber","year":"2005","journal-title":"Phys. Rev. B"},{"first-page":"p 696","year":"1988","author":"Rumelhart","key":"mlstaba821bib48"},{"first-page":"p 65","year":"1992","author":"Hecht-Nielsen","key":"mlstaba821bib49"},{"key":"mlstaba821bib50","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevLett.115.230501","volume":"115","author":"Steiger","year":"2015","journal-title":"Phys. Rev. Lett."},{"journal-title":"Machine learning generated configurations in presence of a conserved quantity: a cautionary tale","year":"2020","author":"Azizi","key":"mlstaba821bib51"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821","content-type":"text\/html","content-version":"vor","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,8,10]],"date-time":"2024-08-10T13:55:45Z","timestamp":1723298145000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/aba821"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,9,11]]},"references-count":51,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2020,9,11]]},"published-print":{"date-parts":[[2020,12,1]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/aba821","relation":{},"ISSN":["2632-2153"],"issn-type":[{"type":"electronic","value":"2632-2153"}],"subject":[],"published":{"date-parts":[[2020,9,11]]},"assertion":[{"value":"Machine learning in physics: the pitfalls of poisoned training sets","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2020 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2020-03-17","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2020-07-21","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2020-09-11","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}