{"id":"https://openalex.org/W2995773176","doi":null,"title":"Implicit Bias of Gradient Descent based Adversarial Training on Separable Data","display_name":"Implicit Bias of Gradient Descent based Adversarial Training on Separable Data","publication_year":2020,"publication_date":"2020-04-30","ids":{"openalex":"https://openalex.org/W2995773176","mag":"2995773176"},"language":"en","primary_location":{"id":"mag:2995773176","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/1906.02931.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/1906.02931.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100380266","display_name":"Yan Li","orcid":"https://orcid.org/0009-0009-3154-4821","observed_orcids":["https://orcid.org/0009-0009-3154-4821","https://orcid.org/0000-0001-7605-1670"]},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yan Li","raw_affiliation_strings":["#N#\u2021#N#Georgia Institute of Technology#N#"],"raw_orcid":"https://orcid.org/0000-0001-7605-1670","affiliations":[{"raw_affiliation_string":"#N#\u2021#N#Georgia Institute of Technology#N#","institution_ids":["https://openalex.org/I130701444"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011349573","display_name":"Ethan X. Fang","orcid":"https://orcid.org/0000-0003-3762-9155","observed_orcids":["https://orcid.org/0000-0003-3762-9155"]},"institutions":[{"id":"https://openalex.org/I130769515","display_name":"Pennsylvania State University","ror":"https://ror.org/04p491231","country_code":"US","type":"education","lineage":["https://openalex.org/I130769515"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ethan X. Fang","raw_affiliation_strings":["Pennsylvania State University"],"raw_orcid":"https://orcid.org/0000-0003-3762-9155","affiliations":[{"raw_affiliation_string":"Pennsylvania State University","institution_ids":["https://openalex.org/I130769515"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Huan Xu","orcid":"https://orcid.org/0000-0002-4468-5019","observed_orcids":["https://orcid.org/0000-0002-4468-5019"]},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Huan Xu","raw_affiliation_strings":["#N#\u2021#N#Georgia Institute of Technology#N#"],"raw_orcid":"https://orcid.org/0000-0002-4468-5019","affiliations":[{"raw_affiliation_string":"#N#\u2021#N#Georgia Institute of Technology#N#","institution_ids":["https://openalex.org/I130701444"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101595501","display_name":"Tuo Zhao","orcid":"https://orcid.org/0000-0002-4991-7851","observed_orcids":["https://orcid.org/0000-0002-4991-7851"]},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tuo Zhao","raw_affiliation_strings":["#N#\u2021#N#Georgia Institute of Technology#N#"],"raw_orcid":"https://orcid.org/0000-0002-4991-7851","affiliations":[{"raw_affiliation_string":"#N#\u2021#N#Georgia Institute of Technology#N#","institution_ids":["https://openalex.org/I130701444"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":19,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9922000169754028,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9922000169754028,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.005400000140070915,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.0007999999797903001,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/adversarial-training","display_name":"adversarial training","score":0.9670000076293945},{"id":"https://openalex.org/keywords/implicit-bias","display_name":"implicit bias","score":0.8190000057220459},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"gradient descent","score":0.9900000095367432},{"id":"https://openalex.org/keywords/linearly-separable-data","display_name":"linearly separable data","score":0.8270000219345093},{"id":"https://openalex.org/keywords/binary-classification","display_name":"binary classification","score":0.7749999761581421},{"id":"https://openalex.org/keywords/robust-neural-networks","display_name":"robust neural networks","score":0.8029999732971191},{"id":"https://openalex.org/keywords/adversarial-perturbations","display_name":"adversarial perturbations","score":0.8080000281333923},{"id":"https://openalex.org/keywords/convergence-rate","display_name":"convergence rate","score":0.7720000147819519},{"id":"https://openalex.org/keywords/worst-case-robustness","display_name":"worst-case robustness","score":0.6740000247955322}],"concepts":[{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.7052351832389832},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.636526346206665},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.5632451772689819},{"id":"https://openalex.org/C173102733","wikidata":"https://www.wikidata.org/wiki/Q6760396","display_name":"Margin classifier","level":3,"score":0.5115895867347717},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.5072605013847351},{"id":"https://openalex.org/C70710897","wikidata":"https://www.wikidata.org/wiki/Q680081","display_name":"Separable space","level":2,"score":0.4995901584625244},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.49088436365127563},{"id":"https://openalex.org/C177918212","wikidata":"https://www.wikidata.org/wiki/Q803623","display_name":"Perturbation (astronomy)","level":2,"score":0.484607070684433},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.45845240354537964},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.44818970561027527},{"id":"https://openalex.org/C66905080","wikidata":"https://www.wikidata.org/wiki/Q17005494","display_name":"Binary classification","level":3,"score":0.44238877296447754},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4386853873729706},{"id":"https://openalex.org/C149728462","wikidata":"https://www.wikidata.org/wiki/Q751319","display_name":"Minimax","level":2,"score":0.42818811535835266},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.42687004804611206},{"id":"https://openalex.org/C95611797","wikidata":"https://www.wikidata.org/wiki/Q17502105","display_name":"Infimum and supremum","level":2,"score":0.4191363751888275},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3749980926513672},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.34508609771728516},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.31468552350997925},{"id":"https://openalex.org/C118615104","wikidata":"https://www.wikidata.org/wiki/Q121416","display_name":"Discrete mathematics","level":1,"score":0.20064136385917664},{"id":"https://openalex.org/C12267149","wikidata":"https://www.wikidata.org/wiki/Q282453","display_name":"Support vector machine","level":2,"score":0.17109894752502441},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.09361010789871216},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"mag:2995773176","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/1906.02931.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null}],"best_oa_location":{"id":"mag:2995773176","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/1906.02931.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":35,"referenced_works":["https://openalex.org/W577198184","https://openalex.org/W1720114023","https://openalex.org/W1836465849","https://openalex.org/W1899249567","https://openalex.org/W1905905554","https://openalex.org/W2093579479","https://openalex.org/W2095705004","https://openalex.org/W2163605009","https://openalex.org/W2174868984","https://openalex.org/W2184045248","https://openalex.org/W2243397390","https://openalex.org/W2516574342","https://openalex.org/W2517229335","https://openalex.org/W2603766943","https://openalex.org/W2766965791","https://openalex.org/W2787708942","https://openalex.org/W2788067318","https://openalex.org/W2790253170","https://openalex.org/W2912322140","https://openalex.org/W2913266441","https://openalex.org/W2941057241","https://openalex.org/W2950220847","https://openalex.org/W2962872506","https://openalex.org/W2963100491","https://openalex.org/W2963207607","https://openalex.org/W2963208657","https://openalex.org/W2963431851","https://openalex.org/W2963539647","https://openalex.org/W2963655672","https://openalex.org/W2963695615","https://openalex.org/W2963826371","https://openalex.org/W2963837241","https://openalex.org/W2963959597","https://openalex.org/W2964153729","https://openalex.org/W2964253222"],"related_works":["https://openalex.org/W2948531553","https://openalex.org/W7047792466","https://openalex.org/W2945724059","https://openalex.org/W4410950114","https://openalex.org/W4403444251","https://openalex.org/W4289713235","https://openalex.org/W3095929309","https://openalex.org/W4221161843","https://openalex.org/W4378465428","https://openalex.org/W7133243822"],"abstract_inverted_index":{"Adversarial":[0],"training":[1,7,37,92,106,130,141,189],"is":[2],"a":[3,153,160],"principled":[4],"approach":[5],"for":[6,82,187],"robust":[8],"neural":[9],"networks.":[10],"Despite":[11],"of":[12,32,120,129,163],"tremendous":[13],"successes":[14],"in":[15,108,150],"practice,":[16],"its":[17,40,45,67],"theoretical":[18,30,185],"properties":[19],"still":[20],"remain":[21],"largely":[22],"unexplored.":[23],"In":[24,134],"this":[25],"paper,":[26],"we":[27,79],"provide":[28,184],"new":[29],"insights":[31],"gradient":[33,102],"descent":[34,103],"based":[35,104],"adversarial":[36,89,105,138,188,196],"by":[38,101],"studying":[39],"computational":[41],"properties,":[42],"specifically":[43],"on":[44,54],"implicit":[46],"bias.":[47],"We":[48],"take":[49],"the":[50,63,70,88,98,111,118,125,137,146,167,180],"binary":[51],"classification":[52],"task":[53],"linearly":[55],"separable":[56],"data":[57],"as":[58,69,165],"an":[59],"illustrative":[60],"example,":[61],"where":[62],"loss":[64],"asymptotically":[65],"attains":[66],"infimum":[68],"parameter":[71],"diverges":[72],"to":[73,110,152,179],"infinity":[74],"along":[75],"certain":[76],"directions.":[77],"Specifically,":[78],"show":[80],"that":[81,190],"any":[83],"fixed":[84],"iteration":[85],"$T$,":[86],"when":[87,136],"perturbation":[90,139,178],"during":[91,140],"has":[93,142,159],"proper":[94],"bounded":[95,143,175],"L2":[96,113,169],"norm,":[97,145],"classifier":[99,116,148,172],"learned":[100],"converges":[107,149],"direction":[109,151],"maximum":[112,154,168],"norm":[114,170,177],"margin":[115,156,171],"at":[117],"rate":[119,126],"$O(1/\\sqrt{T})$,":[121],"significantly":[122],"faster":[123],"than":[124],"$O(1/\\log":[127],"T}$":[128],"with":[131],"clean":[132],"data.":[133,181],"addition,":[135],"Lq":[144,176],"resulting":[147],"mixed-norm":[155],"classifier,":[157],"which":[158],"natural":[161],"interpretation":[162],"robustness,":[164],"being":[166],"under":[173],"worst-case":[174],"Our":[182],"findings":[183],"backups":[186],"it":[191],"indeed":[192],"promotes":[193],"robustness":[194],"against":[195],"perturbation.":[197]},"counts_by_year":[{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":1},{"year":2022,"cited_by_count":2},{"year":2021,"cited_by_count":10},{"year":2020,"cited_by_count":5}],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2019-12-26T00:00:00"}