{"id":"https://openalex.org/W7154618849","doi":"https://doi.org/10.48550/arxiv.2604.13870","title":"Gradient Descent's Last Iterate is Often (slightly) Suboptimal","display_name":"Gradient Descent's Last Iterate is Often (slightly) Suboptimal","publication_year":2026,"publication_date":"2026-04-15","ids":{"openalex":"https://openalex.org/W7154618849","doi":"https://doi.org/10.48550/arxiv.2604.13870"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.13870","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13870","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.13870","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5064469637","display_name":"Guy Kornowski","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kornowski, Guy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133811900","display_name":"Ohad Shamir","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shamir, Ohad","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"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":0.9692000150680542,"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":0.9692000150680542,"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/T12056","display_name":"Markov Chains and Monte Carlo Methods","score":0.010099999606609344,"subfield":{"id":"https://openalex.org/subfields/2613","display_name":"Statistics and Probability"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10500","display_name":"Sparse and Compressive Sensing Techniques","score":0.004000000189989805,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.6991999745368958},{"id":"https://openalex.org/keywords/lipschitz-continuity","display_name":"Lipschitz continuity","score":0.682200014591217},{"id":"https://openalex.org/keywords/convex-function","display_name":"Convex function","score":0.5001000165939331},{"id":"https://openalex.org/keywords/convergence","display_name":"Convergence (economics)","score":0.4927000105381012},{"id":"https://openalex.org/keywords/function","display_name":"Function (biology)","score":0.48159998655319214},{"id":"https://openalex.org/keywords/regular-polygon","display_name":"Regular polygon","score":0.45809999108314514},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.42669999599456787}],"concepts":[{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.6991999745368958},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.6947000026702881},{"id":"https://openalex.org/C22324862","wikidata":"https://www.wikidata.org/wiki/Q652707","display_name":"Lipschitz continuity","level":2,"score":0.682200014591217},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.5133000016212463},{"id":"https://openalex.org/C145446738","wikidata":"https://www.wikidata.org/wiki/Q319913","display_name":"Convex function","level":3,"score":0.5001000165939331},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.4927000105381012},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.48159998655319214},{"id":"https://openalex.org/C112680207","wikidata":"https://www.wikidata.org/wiki/Q714886","display_name":"Regular polygon","level":2,"score":0.45809999108314514},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.42669999599456787},{"id":"https://openalex.org/C57869625","wikidata":"https://www.wikidata.org/wiki/Q1783502","display_name":"Rate of convergence","level":3,"score":0.40959998965263367},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.39169999957084656},{"id":"https://openalex.org/C99414536","wikidata":"https://www.wikidata.org/wiki/Q7098950","display_name":"Optimal stopping","level":2,"score":0.37959998846054077},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.3481999933719635},{"id":"https://openalex.org/C71287196","wikidata":"https://www.wikidata.org/wiki/Q847204","display_name":"Limit of a sequence","level":3,"score":0.30169999599456787},{"id":"https://openalex.org/C157972887","wikidata":"https://www.wikidata.org/wiki/Q463359","display_name":"Convex optimization","level":3,"score":0.29670000076293945},{"id":"https://openalex.org/C2776637919","wikidata":"https://www.wikidata.org/wiki/Q624380","display_name":"Descent (aeronautics)","level":2,"score":0.28940001130104065},{"id":"https://openalex.org/C99888217","wikidata":"https://www.wikidata.org/wiki/Q1288707","display_name":"Stopping time","level":2,"score":0.28929999470710754},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.2720000147819519}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.13870","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13870","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.13870","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13870","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"We":[0,118],"consider":[1],"the":[2,23,58,104,129],"well-studied":[3],"setting":[4],"of":[5,43,52,97,132],"minimizing":[6],"a":[7,38,64,111],"convex":[8],"Lipschitz":[9],"function":[10],"using":[11],"either":[12],"gradient":[13],"descent":[14],"(GD)":[15],"or":[16],"its":[17],"stochastic":[18],"variant":[19],"(SGD),":[20],"and":[21,122],"examine":[22],"last":[24,39,109,149],"iterate":[25,40,150],"convergence.":[26],"By":[27],"now,":[28],"it":[29,134],"is":[30,135],"known":[31],"that":[32,93,126,156],"standard":[33],"stepsize":[34,66,80,100],"choices":[35],"lead":[36],"to":[37,78,137],"convergence":[41],"rate":[42,61],"$\\log":[44],"T/\\sqrt{T}$":[45],"after":[46],"$T$":[47,73,144],"steps.":[48],"A":[49],"breakthrough":[50],"result":[51],"Jain":[53,89],"et":[54,90],"al.":[55,91],"[2019]":[56],"recovered":[57],"optimal":[59,105],"$1/\\sqrt{T}$":[60],"by":[62],"constructing":[63],"non-standard":[65],"sequence.":[67],"However,":[68],"this":[69,120],"sequence":[70,101],"requires":[71],"choosing":[72],"in":[74,123,128,143],"advance,":[75],"as":[76],"opposed":[77],"common":[79],"schedules":[81],"which":[82,113],"apply":[83],"for":[84,107],"any":[85],"time":[86],"horizon.":[87],"Moreover,":[88],"conjectured":[92],"without":[94],"prior":[95],"knowledge":[96],"$T$,":[98],"no":[99],"can":[102],"ensure":[103],"error":[106],"SGD's":[108],"iterate,":[110],"claim":[112],"so":[114],"far":[115],"remained":[116],"unproven.":[117],"prove":[119],"conjecture,":[121],"fact":[124],"show":[125],"even":[127],"noiseless":[130],"case":[131],"GD,":[133],"impossible":[136],"avoid":[138],"an":[139,147],"excess":[140],"poly-log":[141],"factor":[142],"when":[145],"considering":[146],"anytime":[148],"guarantee.":[151],"Our":[152],"proof":[153],"further":[154],"suggests":[155],"such":[157],"(slightly)":[158],"suboptimal":[159],"stopping":[160],"times":[161],"are":[162],"unavoidably":[163],"common.":[164]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-17T00:00:00"}