{"id":"https://openalex.org/W4401066014","doi":"https://doi.org/10.48550/arxiv.2402.18752","title":"Pre-training Differentially Private Models with Limited Public Data","display_name":"Pre-training Differentially Private Models with Limited Public Data","publication_year":2024,"publication_date":"2024-02-28","ids":{"openalex":"https://openalex.org/W4401066014","doi":"https://doi.org/10.48550/arxiv.2402.18752"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2402.18752","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2402.18752","pdf_url":"https://arxiv.org/pdf/2402.18752","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":null},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2402.18752","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5109754291","display_name":"Zhiqi Bu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bu, Zhiqi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5106488031","display_name":"Xinwei Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xinwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100633783","display_name":"Mingyi Hong","orcid":"https://orcid.org/0000-0003-1263-9365"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hong, Mingyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5025057192","display_name":"Sheng Zha","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zha, Sheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5082384108","display_name":"George Karypis","orcid":"https://orcid.org/0000-0003-2753-1437"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Karypis, George","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":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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.9872999787330627,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.9872999787330627,"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/T11182","display_name":"Auction Theory and Applications","score":0.9861999750137329,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.6497965455055237},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.3664276599884033},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.3265117406845093},{"id":"https://openalex.org/keywords/geography","display_name":"Geography","score":0.06302139163017273}],"concepts":[{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.6497965455055237},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3664276599884033},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.3265117406845093},{"id":"https://openalex.org/C205649164","wikidata":"https://www.wikidata.org/wiki/Q1071","display_name":"Geography","level":0,"score":0.06302139163017273},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2402.18752","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2402.18752","pdf_url":"https://arxiv.org/pdf/2402.18752","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":null},{"id":"doi:10.48550/arxiv.2402.18752","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2402.18752","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":"pmh:oai:arXiv.org:2402.18752","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2402.18752","pdf_url":"https://arxiv.org/pdf/2402.18752","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":null},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320309636","display_name":"University of Minnesota","ror":"https://ror.org/03grvy078"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W230091440","https://openalex.org/W2390279801","https://openalex.org/W2233261550","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2810751659","https://openalex.org/W258997015","https://openalex.org/W2376932109"],"abstract_inverted_index":{"The":[0],"superior":[1],"performance":[2,60,118],"of":[3,11,14,41,75,80,98,101,127,144,153,164],"large":[4],"foundation":[5],"models":[6,192],"relies":[7],"on":[8,155,169,176],"the":[9,39,45,53,59,66,81,85,99,106,125],"use":[10,126],"massive":[12],"amounts":[13],"high-quality":[15],"data,":[16,130,146],"which":[17,131],"often":[18],"contain":[19],"sensitive,":[20],"private":[21],"and":[22,166,167,173,182],"copyrighted":[23],"material":[24],"that":[25,115],"requires":[26],"formal":[27],"protection.":[28],"While":[29],"differential":[30],"privacy":[31],"(DP)":[32],"is":[33,49,71],"a":[34,77,95,112,134],"prominent":[35],"method":[36],"to":[37,44,52,58,133],"gauge":[38],"degree":[40],"security":[42],"provided":[43],"models,":[46],"its":[47],"application":[48],"commonly":[50],"limited":[51,128],"model":[54],"fine-tuning":[55],"stage,":[56],"due":[57],"degradation":[61,119],"when":[62],"applying":[63],"DP":[64,70,102,116,136,151,186,190],"during":[65,84],"pre-training":[67,87,138,181],"stage.":[68],"Consequently,":[69],"yet":[72],"not":[73],"capable":[74],"protecting":[76],"substantial":[78],"portion":[79],"data":[82],"used":[83],"initial":[86],"process.":[88],"In":[89],"this":[90],"work,":[91],"we":[92],"first":[93],"provide":[94],"theoretical":[96],"understanding":[97],"efficacy":[100],"training":[103],"by":[104,124],"analyzing":[105],"per-iteration":[107],"loss":[108],"improvement.":[109],"We":[110],"make":[111],"key":[113],"observation":[114],"optimizers'":[117],"can":[120,149],"be":[121],"significantly":[122],"mitigated":[123],"public":[129,145],"leads":[132],"novel":[135],"continual":[137],"strategy.":[139],"Empirically,":[140],"using":[141],"only":[142],"10\\%":[143],"our":[147],"strategy":[148],"achieve":[150],"accuracy":[152,163],"41.5\\%":[154],"ImageNet-21k":[156],"(with":[157],"$\u03b5=8$),":[158],"as":[159,161],"well":[160],"non-DP":[162],"55.7\\%":[165],"60.0\\%":[168],"downstream":[170],"tasks":[171],"Places365":[172],"iNaturalist-2021,":[174],"respectively,":[175],"par":[177],"with":[178],"state-of-the-art":[179],"standard":[180],"substantially":[183],"outperforming":[184],"existing":[185],"pre-trained":[187,191],"models.":[188],"Our":[189],"are":[193],"released":[194],"in":[195],"fastDP":[196],"library":[197],"(https://github.com/awslabs/fast-differential-privacy/releases/tag/v2.1)":[198]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2025-10-10T00:00:00"}