{"id":"https://openalex.org/W2993367001","doi":"https://doi.org/10.1109/allerton.2019.8919696","title":"Gradient Descent Finds Global Minima for Generalizable Deep Neural Networks of Practical Sizes","display_name":"Gradient Descent Finds Global Minima for Generalizable Deep Neural Networks of Practical Sizes","publication_year":2019,"publication_date":"2019-09-01","ids":{"openalex":"https://openalex.org/W2993367001","doi":"https://doi.org/10.1109/allerton.2019.8919696","mag":"2993367001"},"language":"en","primary_location":{"id":"doi:10.1109/allerton.2019.8919696","is_oa":false,"landing_page_url":"https://doi.org/10.1109/allerton.2019.8919696","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2019 57th Annual Allerton Conference on Communication, Control, and Computing (Allerton)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5003184366","display_name":"Kenji Kawaguchi","orcid":"https://orcid.org/0000-0002-5361-9793"},"institutions":[{"id":"https://openalex.org/I2801851002","display_name":"Harvard University Press","ror":"https://ror.org/006v7bf86","country_code":"US","type":"other","lineage":["https://openalex.org/I136199984","https://openalex.org/I2801851002"]},{"id":"https://openalex.org/I4210109586","display_name":"Moscow Institute of Thermal Technology","ror":"https://ror.org/021es5e59","country_code":"RU","type":"facility","lineage":["https://openalex.org/I4210109586"]}],"countries":["RU","US"],"is_corresponding":true,"raw_author_name":"Kenji Kawaguchi","raw_affiliation_strings":["MIT","Harvard University"],"affiliations":[{"raw_affiliation_string":"MIT","institution_ids":["https://openalex.org/I4210109586"]},{"raw_affiliation_string":"Harvard University","institution_ids":["https://openalex.org/I2801851002"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5007414666","display_name":"Jiaoyang Huang","orcid":"https://orcid.org/0000-0001-8266-1301"},"institutions":[{"id":"https://openalex.org/I2801851002","display_name":"Harvard University Press","ror":"https://ror.org/006v7bf86","country_code":"US","type":"other","lineage":["https://openalex.org/I136199984","https://openalex.org/I2801851002"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jiaoyang Huang","raw_affiliation_strings":["Harvard University"],"affiliations":[{"raw_affiliation_string":"Harvard University","institution_ids":["https://openalex.org/I2801851002"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5003184366"],"corresponding_institution_ids":["https://openalex.org/I2801851002","https://openalex.org/I4210109586"],"apc_list":null,"apc_paid":null,"fwci":3.8964,"has_fulltext":false,"cited_by_count":43,"citation_normalized_percentile":{"value":0.95000996,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"92","last_page":"99"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9994999766349792,"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":0.9994999766349792,"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":0.9972000122070312,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.9965999722480774,"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/maxima-and-minima","display_name":"Maxima and minima","score":0.8276236057281494},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.7777484059333801},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.7135280966758728},{"id":"https://openalex.org/keywords/logarithm","display_name":"Logarithm","score":0.7032657265663147},{"id":"https://openalex.org/keywords/deep-neural-networks","display_name":"Deep neural networks","score":0.6780844926834106},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.590064287185669},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5856492519378662},{"id":"https://openalex.org/keywords/nonlinear-system","display_name":"Nonlinear system","score":0.5239337086677551},{"id":"https://openalex.org/keywords/stochastic-gradient-descent","display_name":"Stochastic gradient descent","score":0.5025885105133057},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.49058809876441956},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3876647651195526},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.35752108693122864},{"id":"https://openalex.org/keywords/mathematical-analysis","display_name":"Mathematical analysis","score":0.08124411106109619},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.06713283061981201}],"concepts":[{"id":"https://openalex.org/C186633575","wikidata":"https://www.wikidata.org/wiki/Q845060","display_name":"Maxima and minima","level":2,"score":0.8276236057281494},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.7777484059333801},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.7135280966758728},{"id":"https://openalex.org/C39927690","wikidata":"https://www.wikidata.org/wiki/Q11197","display_name":"Logarithm","level":2,"score":0.7032657265663147},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.6780844926834106},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.590064287185669},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5856492519378662},{"id":"https://openalex.org/C158622935","wikidata":"https://www.wikidata.org/wiki/Q660848","display_name":"Nonlinear system","level":2,"score":0.5239337086677551},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.5025885105133057},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49058809876441956},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3876647651195526},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.35752108693122864},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.08124411106109619},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.06713283061981201},{"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":"doi:10.1109/allerton.2019.8919696","is_oa":false,"landing_page_url":"https://doi.org/10.1109/allerton.2019.8919696","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2019 57th Annual Allerton Conference on Communication, Control, and Computing (Allerton)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":47,"referenced_works":["https://openalex.org/W1608733719","https://openalex.org/W2012903341","https://openalex.org/W2112796928","https://openalex.org/W2128865943","https://openalex.org/W2161278885","https://openalex.org/W2166915851","https://openalex.org/W2167982865","https://openalex.org/W2169976759","https://openalex.org/W2181056828","https://openalex.org/W2215006509","https://openalex.org/W2223846163","https://openalex.org/W2565538933","https://openalex.org/W2804822090","https://openalex.org/W2886067286","https://openalex.org/W2899748887","https://openalex.org/W2899790086","https://openalex.org/W2900959181","https://openalex.org/W2947075492","https://openalex.org/W2948046738","https://openalex.org/W2949804919","https://openalex.org/W2950618860","https://openalex.org/W2950987997","https://openalex.org/W2952574409","https://openalex.org/W2962698540","https://openalex.org/W2963759574","https://openalex.org/W2964098911","https://openalex.org/W2964161337","https://openalex.org/W2964232029","https://openalex.org/W2970443625","https://openalex.org/W2971120029","https://openalex.org/W2980775160","https://openalex.org/W3104627075","https://openalex.org/W4289436733","https://openalex.org/W4291417645","https://openalex.org/W6608025168","https://openalex.org/W6636441265","https://openalex.org/W6681017033","https://openalex.org/W6688493498","https://openalex.org/W6731278065","https://openalex.org/W6751437525","https://openalex.org/W6753918066","https://openalex.org/W6755150206","https://openalex.org/W6756001544","https://openalex.org/W6756137178","https://openalex.org/W6763113216","https://openalex.org/W6763304541","https://openalex.org/W6763681727"],"related_works":["https://openalex.org/W4366280654","https://openalex.org/W4362706668","https://openalex.org/W2479609865","https://openalex.org/W3128462138","https://openalex.org/W3103375947","https://openalex.org/W4224998919","https://openalex.org/W4283815402","https://openalex.org/W2993367001","https://openalex.org/W2742801844","https://openalex.org/W2860351735"],"abstract_inverted_index":{"In":[0],"this":[1,29],"paper,":[2],"we":[3,90],"theoretically":[4],"prove":[5,91],"that":[6,83,92,107],"gradient":[7],"descent":[8],"can":[9],"find":[10],"a":[11,114,135,140],"global":[12],"minimum":[13],"for":[14],"nonlinear":[15],"deep":[16,66,118],"neural":[17,67,119],"networks":[18,68,120],"of":[19,36,47,56,64,79,96,99],"sizes":[20],"commonly":[21],"encountered":[22],"in":[23,28],"practice.":[24],"The":[25],"theory":[26,42],"developed":[27],"paper":[30],"only":[31,43],"requires":[32,44],"the":[33,45,54,62,65,86,93,97,100,103,122],"practical":[34],"degrees":[35],"over-parameterization":[37],"unlike":[38],"previous":[39,87],"theories.":[40,88],"Our":[41],"number":[46,55],"trainable":[48],"parameters":[49],"to":[50,69,75,127,130],"increase":[51,95],"linearly":[52],"as":[53],"training":[57],"samples":[58,133],"increases.":[59],"This":[60],"allows":[61],"size":[63,98],"be":[70,76,110],"consistent":[71],"with":[72,121,134],"practice":[73],"and":[74,106],"several":[77],"orders":[78],"magnitude":[80],"smaller":[81],"than":[82],"required":[84],"by":[85,113],"Moreover,":[89],"linear":[94],"network":[101],"is":[102],"optimal":[104],"rate":[105],"it":[108],"cannot":[109],"improved,":[111],"except":[112],"logarithmic":[115],"factor.":[116],"Furthermore,":[117],"trainability":[123],"guarantee":[124],"are":[125],"shown":[126],"generalize":[128],"well":[129],"unseen":[131],"test":[132],"natural":[136],"dataset":[137],"but":[138],"not":[139],"random":[141],"dataset.":[142]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":7},{"year":2024,"cited_by_count":5},{"year":2023,"cited_by_count":2},{"year":2022,"cited_by_count":2},{"year":2021,"cited_by_count":14},{"year":2020,"cited_by_count":5},{"year":2019,"cited_by_count":6}],"updated_date":"2026-04-02T13:48:15.688549","created_date":"2025-10-10T00:00:00"}