{"id":"https://openalex.org/W4283816219","doi":"https://doi.org/10.48550/arxiv.2207.00160","title":"When Does Differentially Private Learning Not Suffer in High Dimensions?","display_name":"When Does Differentially Private Learning Not Suffer in High Dimensions?","publication_year":2022,"publication_date":"2022-07-01","ids":{"openalex":"https://openalex.org/W4283816219","doi":"https://doi.org/10.48550/arxiv.2207.00160"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2207.00160","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2207.00160","pdf_url":"https://arxiv.org/pdf/2207.00160","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":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2207.00160","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100752866","display_name":"Xuechen Li","orcid":"https://orcid.org/0000-0002-7236-9582"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Li, Xuechen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008447585","display_name":"Daogao Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Daogao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015518638","display_name":"Tatsunori Hashimoto","orcid":"https://orcid.org/0000-0003-0521-5855"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hashimoto, Tatsunori","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071827392","display_name":"Huseyin A. Inan","orcid":"https://orcid.org/0000-0001-7465-7214"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Inan, Huseyin A.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043137436","display_name":"Janardhan Kulkarni","orcid":"https://orcid.org/0009-0005-9980-7836"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kulkarni, Janardhan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024980585","display_name":"Yin Tat Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Yin Tat","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5008213441","display_name":"Abhradeep Thakurta","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thakurta, Abhradeep Guha","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5100752866"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":8,"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":0.9993000030517578,"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":0.9993000030517578,"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/T12229","display_name":"Pharmacological Effects and Toxicity Studies","score":0.9771000146865845,"subfield":{"id":"https://openalex.org/subfields/2735","display_name":"Pediatrics, Perinatology and Child Health"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.951200008392334,"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/dimension","display_name":"Dimension (graph theory)","score":0.5928742289543152},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5613692998886108},{"id":"https://openalex.org/keywords/regular-polygon","display_name":"Regular polygon","score":0.5464776754379272},{"id":"https://openalex.org/keywords/linear-subspace","display_name":"Linear subspace","score":0.4517795145511627},{"id":"https://openalex.org/keywords/population","display_name":"Population","score":0.4499474763870239},{"id":"https://openalex.org/keywords/factor-analysis","display_name":"Factor analysis","score":0.44255250692367554},{"id":"https://openalex.org/keywords/lipschitz-continuity","display_name":"Lipschitz continuity","score":0.41164571046829224},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.39093464612960815},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3644503653049469},{"id":"https://openalex.org/keywords/econometrics","display_name":"Econometrics","score":0.3397863507270813},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.3384838104248047},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.3160354495048523},{"id":"https://openalex.org/keywords/pure-mathematics","display_name":"Pure mathematics","score":0.1323317587375641}],"concepts":[{"id":"https://openalex.org/C33676613","wikidata":"https://www.wikidata.org/wiki/Q13415176","display_name":"Dimension (graph theory)","level":2,"score":0.5928742289543152},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5613692998886108},{"id":"https://openalex.org/C112680207","wikidata":"https://www.wikidata.org/wiki/Q714886","display_name":"Regular polygon","level":2,"score":0.5464776754379272},{"id":"https://openalex.org/C12362212","wikidata":"https://www.wikidata.org/wiki/Q728435","display_name":"Linear subspace","level":2,"score":0.4517795145511627},{"id":"https://openalex.org/C2908647359","wikidata":"https://www.wikidata.org/wiki/Q2625603","display_name":"Population","level":2,"score":0.4499474763870239},{"id":"https://openalex.org/C10879293","wikidata":"https://www.wikidata.org/wiki/Q726474","display_name":"Factor analysis","level":2,"score":0.44255250692367554},{"id":"https://openalex.org/C22324862","wikidata":"https://www.wikidata.org/wiki/Q652707","display_name":"Lipschitz continuity","level":2,"score":0.41164571046829224},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39093464612960815},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3644503653049469},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.3397863507270813},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3384838104248047},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3160354495048523},{"id":"https://openalex.org/C202444582","wikidata":"https://www.wikidata.org/wiki/Q837863","display_name":"Pure mathematics","level":1,"score":0.1323317587375641},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0},{"id":"https://openalex.org/C149923435","wikidata":"https://www.wikidata.org/wiki/Q37732","display_name":"Demography","level":1,"score":0.0},{"id":"https://openalex.org/C144024400","wikidata":"https://www.wikidata.org/wiki/Q21201","display_name":"Sociology","level":0,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2207.00160","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2207.00160","pdf_url":"https://arxiv.org/pdf/2207.00160","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":"text"},{"id":"doi:10.48550/arxiv.2207.00160","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2207.00160","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":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2207.00160","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2207.00160","pdf_url":"https://arxiv.org/pdf/2207.00160","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":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2391321755","https://openalex.org/W3185235544","https://openalex.org/W2949110394","https://openalex.org/W2896134808","https://openalex.org/W4297791327","https://openalex.org/W4289378085","https://openalex.org/W2071055957","https://openalex.org/W4211228991","https://openalex.org/W2353790329","https://openalex.org/W2047973899"],"abstract_inverted_index":{"Large":[0],"pretrained":[1],"models":[2,27],"can":[3,28,182],"be":[4,183],"privately":[5],"fine-tuned":[6],"to":[7,149,178],"achieve":[8,29],"performance":[9,56],"approaching":[10],"that":[11,25,69,81,99,116,125],"of":[12,42,57,72,129],"non-private":[13],"models.":[14],"A":[15],"common":[16],"theme":[17],"in":[18,126,157,173],"these":[19],"results":[20,37,164,181],"is":[21,77,147],"the":[22,39,49,55,70,97,110],"surprising":[23],"observation":[24],"high-dimensional":[26],"favorable":[30],"privacy-utility":[31],"trade-offs.":[32],"This":[33,145],"seemingly":[34],"contradicts":[35],"known":[36],"on":[38,96],"model-size":[40],"dependence":[41],"differentially":[43,58],"private":[44,59,89,127,175],"convex":[45,90,158],"learning":[46,60],"and":[47,105,113,162],"raises":[48],"following":[50],"research":[51],"question:":[52],"When":[53],"does":[54],"not":[61],"degrade":[62],"with":[63],"increasing":[64],"model":[65],"size?":[66],"We":[67,122],"identify":[68],"magnitudes":[71],"gradients":[73,133],"projected":[74],"onto":[75],"subspaces":[76],"a":[78,94,141,167],"key":[79],"factor":[80],"determines":[82],"performance.":[83],"To":[84],"precisely":[85],"characterize":[86],"this":[87],"for":[88,109,170],"learning,":[91],"we":[92,100,153],"introduce":[93],"condition":[95],"objective":[98],"term":[101],"\\emph{restricted":[102],"Lipschitz":[103],"continuity}":[104],"derive":[106],"improved":[107],"bounds":[108,156],"excess":[111],"empirical":[112,163],"population":[114],"risks":[115],"are":[117,137],"dimension-independent":[118,155],"under":[119,151],"additional":[120],"conditions.":[121],"empirically":[123],"show":[124],"fine-tuning":[128,136],"large":[130],"language":[131],"models,":[132],"obtained":[134],"during":[135],"mostly":[138],"controlled":[139],"by":[140],"few":[142],"principal":[143],"components.":[144],"behavior":[146],"similar":[148],"conditions":[150],"which":[152],"obtain":[154],"settings.":[159],"Our":[160],"theoretical":[161],"together":[165],"provide":[166],"possible":[168],"explanation":[169],"recent":[171],"successes":[172],"large-scale":[174],"fine-tuning.":[176],"Code":[177],"reproduce":[179],"our":[180],"found":[184],"at":[185],"\\url{https://github.com/lxuechen/private-transformers/tree/main/examples/classification/spectral_analysis}.":[186]},"counts_by_year":[{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":4},{"year":2023,"cited_by_count":2}],"updated_date":"2026-05-03T08:25:01.440150","created_date":"2025-10-10T00:00:00"}