{"id":"https://openalex.org/W7127915844","doi":"https://doi.org/10.48550/arxiv.2602.03872","title":"Understanding the Impact of Differentially Private Training on Memorization of Long-Tailed Data","display_name":"Understanding the Impact of Differentially Private Training on Memorization of Long-Tailed Data","publication_year":2026,"publication_date":"2026-02-01","ids":{"openalex":"https://openalex.org/W7127915844","doi":"https://doi.org/10.48550/arxiv.2602.03872"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.03872","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"type":"article","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100453782","display_name":"Jiaming Zhang","orcid":"https://orcid.org/0000-0003-0991-7109"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jiaming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102697870","display_name":"Huanyi Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Huanyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125172179","display_name":"Meng Ding","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ding, Meng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125208819","display_name":"Shaopeng Fu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fu, Shaopeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101838647","display_name":"Jinyan Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Jinyan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5125141062","display_name":"Di Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Di","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":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.14702918,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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":0.9754999876022339,"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":0.9754999876022339,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.004100000020116568,"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/T11704","display_name":"Mobile Crowdsensing and Crowdsourcing","score":0.0013000000035390258,"subfield":{"id":"https://openalex.org/subfields/1706","display_name":"Computer Science Applications"},"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/memorization","display_name":"Memorization","score":0.8684999942779541},{"id":"https://openalex.org/keywords/differential-privacy","display_name":"Differential privacy","score":0.6057000160217285},{"id":"https://openalex.org/keywords/differential-privacy","display_name":"Differential privacy","score":0.6055999994277954},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.5282999873161316},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.5016000270843506},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.4593000113964081},{"id":"https://openalex.org/keywords/affect","display_name":"Affect (linguistics)","score":0.41920000314712524},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.39430001378059387},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.38269999623298645}],"concepts":[{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.8684999942779541},{"id":"https://openalex.org/C23130292","wikidata":"https://www.wikidata.org/wiki/Q5275358","display_name":"Differential privacy","level":2,"score":0.6057000160217285},{"id":"https://openalex.org/C23130292","wikidata":"https://www.wikidata.org/wiki/Q5275358","display_name":"Differential privacy","level":2,"score":0.6055999994277954},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6031000018119812},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5501000285148621},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5468000173568726},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.5282999873161316},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5016000270843506},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.4593000113964081},{"id":"https://openalex.org/C2776035688","wikidata":"https://www.wikidata.org/wiki/Q1606558","display_name":"Affect (linguistics)","level":2,"score":0.41920000314712524},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.39430001378059387},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.38269999623298645},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.3822999894618988},{"id":"https://openalex.org/C16910744","wikidata":"https://www.wikidata.org/wiki/Q7705759","display_name":"Test data","level":2,"score":0.3790999948978424},{"id":"https://openalex.org/C2776848632","wikidata":"https://www.wikidata.org/wiki/Q853463","display_name":"Clipping (morphology)","level":2,"score":0.36230000853538513},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.36070001125335693},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.33980000019073486},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.3312000036239624},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3303000032901764},{"id":"https://openalex.org/C87007009","wikidata":"https://www.wikidata.org/wiki/Q210832","display_name":"Statistical hypothesis testing","level":2,"score":0.2831999957561493},{"id":"https://openalex.org/C93226319","wikidata":"https://www.wikidata.org/wiki/Q193137","display_name":"Differential (mechanical device)","level":2,"score":0.2824999988079071},{"id":"https://openalex.org/C181447626","wikidata":"https://www.wikidata.org/wiki/Q1224359","display_name":"Differential item functioning","level":4,"score":0.28040000796318054},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2791000008583069},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C198477413","wikidata":"https://www.wikidata.org/wiki/Q7647069","display_name":"Survey data collection","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C55037315","wikidata":"https://www.wikidata.org/wiki/Q5421151","display_name":"Experimental data","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.26010000705718994},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.25859999656677246},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.03872","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2602.03872","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.03872","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.03872","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"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":{"Recent":[0],"research":[1],"shows":[2,43],"that":[3,44,56,120],"modern":[4],"deep":[5],"learning":[6,116],"models":[7,126],"achieve":[8],"high":[9],"predictive":[10],"accuracy":[11],"partly":[12],"by":[13],"memorizing":[14],"individual":[15],"training":[16,31,148],"samples.":[17,65,170],"Such":[18],"memorization":[19],"raises":[20],"serious":[21],"privacy":[22,81],"concerns,":[23],"motivating":[24],"the":[25,88,103,121,128,135,140,147,162],"widespread":[26],"adoption":[27],"of":[28,40,61,72,124,150],"differentially":[29],"private":[30],"algorithms":[32],"such":[33],"as":[34],"DP-SGD.":[35],"However,":[36],"a":[37,58,69,114],"growing":[38],"body":[39],"empirical":[41],"work":[42],"DP-SGD":[45,109],"often":[46],"leads":[47],"to":[48,85,87,165],"suboptimal":[49],"generalization":[50],"performance,":[51],"particularly":[52],"on":[53,110,127,180],"long-tailed":[54,111,129],"data":[55,112],"contain":[57],"large":[59],"number":[60],"rare":[62],"or":[63],"atypical":[64],"Despite":[66],"these":[67],"observations,":[68],"theoretical":[70,105,175],"understanding":[71],"this":[73,99],"phenomenon":[74],"remains":[75],"largely":[76],"unexplored,":[77],"and":[78,90,156,183],"existing":[79],"differential":[80],"analysis":[82,144],"are":[83],"difficult":[84],"extend":[86],"nonconvex":[89],"nonsmooth":[91],"neural":[92],"networks":[93],"commonly":[94],"used":[95],"in":[96],"practice.":[97],"In":[98],"work,":[100],"we":[101,172],"develop":[102],"first":[104],"framework":[106],"for":[107],"analyzing":[108],"from":[113],"feature":[115],"perspective.":[117],"We":[118],"show":[119],"test":[122,137],"error":[123,138],"DP-SGD-trained":[125],"subpopulation":[130],"is":[131],"significantly":[132],"larger":[133],"than":[134],"overall":[136],"over":[139],"entire":[141],"dataset.":[142],"Our":[143],"further":[145],"characterizes":[146],"dynamics":[149],"DP-SGD,":[151],"demonstrating":[152],"how":[153],"gradient":[154],"clipping":[155],"noise":[157],"injection":[158],"jointly":[159],"adversely":[160],"affect":[161],"model's":[163],"ability":[164],"memorize":[166],"informative":[167],"but":[168],"underrepresented":[169],"Finally,":[171],"validate":[173],"our":[174],"findings":[176],"through":[177],"extensive":[178],"experiments":[179],"both":[181],"synthetic":[182],"real-world":[184],"datasets.":[185]},"counts_by_year":[],"updated_date":"2026-07-15T18:14:33.161393","created_date":"2026-02-07T00:00:00"}