{"id":"https://openalex.org/W7223069161","doi":"https://doi.org/10.48550/arxiv.2610.26748","title":"Gradient descent reaches arbitrary accuracy in a fixed number of large steps on multiclass logistic regression","display_name":"Gradient descent reaches arbitrary accuracy in a fixed number of large steps on multiclass logistic regression","publication_year":2026,"publication_date":"2026-10-19","ids":{"openalex":"https://openalex.org/W7223069161","doi":"https://doi.org/10.48550/arxiv.2610.26748"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.26748","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.26748","pdf_url":"https://arxiv.org/pdf/2610.26748","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":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.26748","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5120440662","display_name":"Andreea-Alexandra Mu\u015fat","orcid":"https://orcid.org/0009-0007-3737-7679","observed_orcids":["https://orcid.org/0009-0007-3737-7679"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Musat, Andreea-Alexandra","raw_affiliation_strings":["{andreea.musat ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"{andreea.musat ,","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5085231507","display_name":"Nicolas Boumal","orcid":"https://orcid.org/0000-0002-1322-958X","observed_orcids":["https://orcid.org/0000-0002-1322-958X"]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Boumal, Nicolas","raw_affiliation_strings":["\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL) , Institute of Mathematics ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne (EPFL) , Institute of Mathematics ,","institution_ids":["https://openalex.org/I5124864"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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":0.9980000257492065,"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":0.9980000257492065,"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":0.0006000000284984708,"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":0.0003000000142492354,"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/gradient-descent","display_name":"gradient descent","score":0.9319999814033508},{"id":"https://openalex.org/keywords/multiclass-logistic-regression","display_name":"multiclass logistic regression","score":0.9789999723434448},{"id":"https://openalex.org/keywords/large-step-size","display_name":"large step size","score":0.8320000171661377},{"id":"https://openalex.org/keywords/perceptron-algorithm","display_name":"perceptron algorithm","score":0.593999981880188},{"id":"https://openalex.org/keywords/linear-separability","display_name":"linear separability","score":0.7300000190734863},{"id":"https://openalex.org/keywords/margins","display_name":"margins","score":0.6370000243186951},{"id":"https://openalex.org/keywords/convergence-rate","display_name":"convergence rate","score":0.382999986410141},{"id":"https://openalex.org/keywords/oscillatory-dynamics","display_name":"oscillatory dynamics","score":0.5070000290870667},{"id":"https://openalex.org/keywords/accelerated-optimization","display_name":"accelerated optimization","score":0.35600000619888306}],"concepts":[{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.8226000070571899},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.7202000021934509},{"id":"https://openalex.org/C70710897","wikidata":"https://www.wikidata.org/wiki/Q680081","display_name":"Separable space","level":2,"score":0.6470000147819519},{"id":"https://openalex.org/C33676613","wikidata":"https://www.wikidata.org/wiki/Q13415176","display_name":"Dimension (graph theory)","level":2,"score":0.592199981212616},{"id":"https://openalex.org/C108710211","wikidata":"https://www.wikidata.org/wiki/Q11538","display_name":"Mathematical proof","level":2,"score":0.4763000011444092},{"id":"https://openalex.org/C151956035","wikidata":"https://www.wikidata.org/wiki/Q1132755","display_name":"Logistic regression","level":2,"score":0.46860000491142273},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.4560000002384186},{"id":"https://openalex.org/C60908668","wikidata":"https://www.wikidata.org/wiki/Q690207","display_name":"Perceptron","level":3,"score":0.427700012922287},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.4124000072479248},{"id":"https://openalex.org/C2777027219","wikidata":"https://www.wikidata.org/wiki/Q1284190","display_name":"Constant (computer programming)","level":2,"score":0.3937000036239624},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.37599998712539673},{"id":"https://openalex.org/C83546350","wikidata":"https://www.wikidata.org/wiki/Q1139051","display_name":"Regression","level":2,"score":0.3756999969482422},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3361999988555908},{"id":"https://openalex.org/C41341539","wikidata":"https://www.wikidata.org/wiki/Q7540242","display_name":"Sliced inverse regression","level":3,"score":0.3346000015735626},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.33239999413490295},{"id":"https://openalex.org/C2988035592","wikidata":"https://www.wikidata.org/wiki/Q208042","display_name":"Regression function","level":3,"score":0.3262999951839447},{"id":"https://openalex.org/C17418463","wikidata":"https://www.wikidata.org/wiki/Q3455874","display_name":"Isotonic regression","level":3,"score":0.32030001282691956},{"id":"https://openalex.org/C119322782","wikidata":"https://www.wikidata.org/wiki/Q2662236","display_name":"VC dimension","level":2,"score":0.3190999925136566},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.3149999976158142},{"id":"https://openalex.org/C48921125","wikidata":"https://www.wikidata.org/wiki/Q10861030","display_name":"Linear regression","level":2,"score":0.3028999865055084},{"id":"https://openalex.org/C21080849","wikidata":"https://www.wikidata.org/wiki/Q13611879","display_name":"Data point","level":2,"score":0.27239999175071716},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2721000015735626},{"id":"https://openalex.org/C134466208","wikidata":"https://www.wikidata.org/wiki/Q1520657","display_name":"Weight function","level":2,"score":0.2711000144481659},{"id":"https://openalex.org/C98458673","wikidata":"https://www.wikidata.org/wiki/Q917657","display_name":"Step function","level":2,"score":0.2669000029563904},{"id":"https://openalex.org/C22324862","wikidata":"https://www.wikidata.org/wiki/Q652707","display_name":"Lipschitz continuity","level":2,"score":0.257999986410141},{"id":"https://openalex.org/C122041747","wikidata":"https://www.wikidata.org/wiki/Q838611","display_name":"Ball (mathematics)","level":2,"score":0.25609999895095825}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.26748","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.26748","pdf_url":"https://arxiv.org/pdf/2610.26748","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":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.26748","doi":"https://doi.org/10.48550/arxiv.2610.26748","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.26748","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":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.26748","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.26748","pdf_url":"https://arxiv.org/pdf/2610.26748","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":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321806","display_name":"Lunds Universitet","ror":"https://ror.org/012a77v79"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7223069161.pdf","grobid_xml":"https://content.openalex.org/works/W7223069161.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W3009686669","https://openalex.org/W3132026801","https://openalex.org/W4300980664","https://openalex.org/W4399511139","https://openalex.org/W4415109397","https://openalex.org/W7166861958","https://openalex.org/W7167697467","https://openalex.org/W7216454620","https://openalex.org/W7220697009"],"related_works":["https://openalex.org/W7220697009","https://openalex.org/W4392224005","https://openalex.org/W4416125235","https://openalex.org/W4409363458","https://openalex.org/W7129085708","https://openalex.org/W4415132702","https://openalex.org/W4399511139","https://openalex.org/W7212007777","https://openalex.org/W7133213864","https://openalex.org/W7216454620"],"abstract_inverted_index":{"We":[0,179],"show":[1],"that":[2,27,61,89,151],"gradient":[3],"descent":[4],"(GD)":[5],"with":[6],"step":[7,36,165],"size":[8,37],"$\u03b7$":[9,38],"can":[10,159],"achieve":[11],"$\\varepsilon$":[12,48],"optimality":[13],"gap":[14],"on":[15,64,103,154],"the":[16,35,52,65,71,74,104,115,119,155],"multiclass":[17],"logistic":[18,156],"regression":[19,157],"loss":[20,158],"in":[21,95,125,136],"a":[22,45,59,122,169,176],"number":[23,75],"of":[24,30,47,51,70,76,182],"iterations":[25],"$T$":[26,57,126],"is":[28,58],"independent":[29],"$\\varepsilon$.":[31],"To":[32],"do":[33],"so,":[34],"must":[39],"be":[40,93,160],"chosen":[41],"large":[42,164],"enough":[43],"as":[44,108],"function":[46],"(and":[49],"properties":[50,69],"linearly":[53],"separable":[54],"dataset),":[55],"while":[56],"constant":[60],"only":[62],"depends":[63],"data.":[66],"The":[67,100],"relevant":[68],"data":[72],"include":[73],"points":[77],"and":[78,83,118],"their":[79],"margin":[80],"(for":[81,87],"$T$),":[82],"one":[84],"additional":[85],"quantity":[86],"$\u03b7$)":[88],"turns":[90],"out":[91],"to":[92,148],"necessary":[94],"dimension":[96,137],"$d":[97,138],"\\geq":[98],"3$.":[99],"proofs":[101],"build":[102],"known":[105],"fact":[106],"that,":[107],"$\u03b7\\to":[109],"\\infty$,":[110],"suitably":[111],"rescaled":[112],"GD":[113,153],"becomes":[114],"perceptron":[116],"algorithm,":[117],"latter":[120],"finds":[121],"perfect":[123],"classifier":[124],"iterations.":[127],"A":[128],"similar":[129],"claim":[130],"was":[131],"shown":[132],"for":[133],"two":[134],"classes":[135],"=":[139],"2$":[140],"using":[141],"other":[142],"proof":[143],"techniques.":[144],"This":[145],"result":[146],"adds":[147],"previous":[149],"works":[150],"showed":[152],"accelerated":[161],"by":[162,175],"taking":[163],"sizes.":[166],"Those":[167],"provoke":[168],"wild,":[170],"oscillatory":[171],"initial":[172],"phase,":[173],"followed":[174],"stable":[177],"phase.":[178],"provide":[180],"results":[181],"this":[183],"(qualitatively":[184],"different)":[185],"nature,":[186],"too.":[187]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-21T00:00:00"}