{"id":"https://openalex.org/W3211267103","doi":null,"title":"The Equilibrium Hypothesis: Rethinking implicit regularization in Deep Neural Networks.","display_name":"The Equilibrium Hypothesis: Rethinking implicit regularization in Deep Neural Networks.","publication_year":2021,"publication_date":"2021-10-22","ids":{"openalex":"https://openalex.org/W3211267103","mag":"3211267103"},"language":"en","primary_location":{"id":"mag:3211267103","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2110.11749.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":"submittedVersion","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/2110.11749.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5035154370","display_name":"Yizhang Lou","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I40120149","display_name":"University of Oxford","ror":"https://ror.org/052gg0110","country_code":"GB","type":"education","lineage":["https://openalex.org/I40120149"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Yizhang Lou","raw_affiliation_strings":["University of Oxford"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Oxford","institution_ids":["https://openalex.org/I40120149"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5061432677","display_name":"Chris Mingard","orcid":"https://orcid.org/0000-0002-4519-9709","observed_orcids":["https://orcid.org/0000-0002-4519-9709"]},"institutions":[{"id":"https://openalex.org/I40120149","display_name":"University of Oxford","ror":"https://ror.org/052gg0110","country_code":"GB","type":"education","lineage":["https://openalex.org/I40120149"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Chris Mingard","raw_affiliation_strings":["University of Oxford"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Oxford","institution_ids":["https://openalex.org/I40120149"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5089509060","display_name":"Soufiane Hayou","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Soufiane Hayou","raw_affiliation_strings":["National University of Singapore,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Singapore,","institution_ids":["https://openalex.org/I165932596"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.6255000233650208,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.6255000233650208,"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/T10320","display_name":"Neural Networks and Applications","score":0.1737000048160553,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.08990000188350677,"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/implicit-regularization","display_name":"implicit regularization","score":0.9459999799728394},{"id":"https://openalex.org/keywords/equilibrium-hypothesis","display_name":"equilibrium hypothesis","score":0.8880000114440918},{"id":"https://openalex.org/keywords/deep-neural-network","display_name":"deep neural network","score":0.6489999890327454},{"id":"https://openalex.org/keywords/generalization","display_name":"generalization","score":0.7570000290870667},{"id":"https://openalex.org/keywords/layer-alignment","display_name":"layer alignment","score":0.531000018119812},{"id":"https://openalex.org/keywords/label-alignment","display_name":"label alignment","score":0.5879999995231628},{"id":"https://openalex.org/keywords/network-depth-and-width","display_name":"network depth and width","score":0.671999990940094}],"concepts":[{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.8429006338119507},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.64583420753479},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6097586154937744},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.573140025138855},{"id":"https://openalex.org/C50335755","wikidata":"https://www.wikidata.org/wiki/Q483247","display_name":"Phenomenon","level":2,"score":0.5213531255722046},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.5004260540008545},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3880195617675781},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.32467854022979736},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.302465558052063},{"id":"https://openalex.org/C111472728","wikidata":"https://www.wikidata.org/wiki/Q9471","display_name":"Epistemology","level":1,"score":0.09645122289657593},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.08852458000183105},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.06987985968589783}],"mesh":[],"locations_count":1,"locations":[{"id":"mag:3211267103","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2110.11749.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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null}],"best_oa_location":{"id":"mag:3211267103","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2110.11749.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":"submittedVersion","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":38,"referenced_works":["https://openalex.org/W1677182931","https://openalex.org/W1904365287","https://openalex.org/W2112796928","https://openalex.org/W2152722485","https://openalex.org/W2163605009","https://openalex.org/W2331143823","https://openalex.org/W2423689290","https://openalex.org/W2552737632","https://openalex.org/W2566079294","https://openalex.org/W2604607385","https://openalex.org/W2622263826","https://openalex.org/W2731468224","https://openalex.org/W2766111679","https://openalex.org/W2766678531","https://openalex.org/W2803921058","https://openalex.org/W2808460953","https://openalex.org/W2809090039","https://openalex.org/W2904014149","https://openalex.org/W2912322140","https://openalex.org/W2912713668","https://openalex.org/W2913473169","https://openalex.org/W2917668703","https://openalex.org/W2946790101","https://openalex.org/W2959995783","https://openalex.org/W2962685794","https://openalex.org/W2962804662","https://openalex.org/W2978031686","https://openalex.org/W3007325053","https://openalex.org/W3034731342","https://openalex.org/W3042925324","https://openalex.org/W3088274487","https://openalex.org/W3101069636","https://openalex.org/W3116019732","https://openalex.org/W3157484000","https://openalex.org/W3158469333","https://openalex.org/W3167509724","https://openalex.org/W3180313059","https://openalex.org/W3205384336"],"related_works":["https://openalex.org/W4286897178","https://openalex.org/W3125148525","https://openalex.org/W3009359749","https://openalex.org/W7160614058","https://openalex.org/W3046476838","https://openalex.org/W3191734561","https://openalex.org/W3089846728","https://openalex.org/W7109040285","https://openalex.org/W3164400695","https://openalex.org/W2940107683"],"abstract_inverted_index":{"Modern":[0],"Deep":[1],"Neural":[2],"Networks":[3],"(DNNs)":[4],"exhibit":[5],"impressive":[6],"generalization":[7],"properties":[8],"on":[9,33],"a":[10],"variety":[11],"of":[12,20],"tasks":[13],"without":[14],"explicit":[15],"regularization,":[16],"suggesting":[17],"the":[18,57,77,89,95,108,111,124],"existence":[19],"hidden":[21],"regularization":[22,37],"effects.":[23],"Recent":[24],"work":[25],"by":[26],"Baratin":[27],"et":[28],"al.":[29],"(2021)":[30],"sheds":[31],"light":[32],"an":[34,64,120],"intriguing":[35],"implicit":[36,65],"effect,":[38],"showing":[39],"that":[40,55,94,97],"some":[41,99],"layers":[42,96],"are":[43,107],"much":[44],"more":[45],"aligned":[46],"with":[47,110,123],"data":[48,115],"labels":[49],"than":[50],"other":[51],"layers.":[52],"This":[53],"suggests":[54],"as":[56],"network":[58],"grows":[59],"in":[60],"depth":[61],"and":[62,86,103],"width,":[63],"layer":[66],"selection":[67],"phenomenon":[68],"occurs":[69],"during":[70],"training.":[71],"In":[72],"this":[73,81],"work,":[74],"we":[75],"provide":[76],"first":[78],"explanation":[79],"for":[80],"alignment":[82,113],"hierarchy.":[83],"We":[84],"introduce":[85],"empirically":[87],"validate":[88],"Equilibrium":[90],"Hypothesis":[91],"which":[92],"states":[93],"achieve":[98],"balance":[100],"between":[101],"forward":[102],"backward":[104],"information":[105],"loss":[106],"ones":[109],"highest":[112],"to":[114],"labels.":[116],"Our":[117],"experiments":[118],"demonstrate":[119],"excellent":[121],"match":[122],"theoretical":[125],"predictions.":[126]},"counts_by_year":[],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2021-11-08T00:00:00"}