{"id":"https://openalex.org/W7162793014","doi":"https://doi.org/10.48550/arxiv.2605.30312","title":"DP-SAPF: Saliency-Aware Parameter Fine-tuning of Public Models for Differentially Private Image Synthesis","display_name":"DP-SAPF: Saliency-Aware Parameter Fine-tuning of Public Models for Differentially Private Image Synthesis","publication_year":2026,"publication_date":"2026-05-28","ids":{"openalex":"https://openalex.org/W7162793014","doi":"https://doi.org/10.48550/arxiv.2605.30312"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.30312","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.30312","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.2605.30312","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137332706","display_name":"Chen Gong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gong, Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137312245","display_name":"Kecen Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Kecen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137383107","display_name":"Zinan Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Zinan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137393468","display_name":"Tianhao Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Tianhao","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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.3497999906539917,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.3497999906539917,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.1412999927997589,"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/T11105","display_name":"Advanced Image Processing Techniques","score":0.07479999959468842,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/salient","display_name":"Salient","score":0.6729999780654907},{"id":"https://openalex.org/keywords/fidelity","display_name":"Fidelity","score":0.6650000214576721},{"id":"https://openalex.org/keywords/image","display_name":"Image (mathematics)","score":0.600600004196167},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.5748000144958496},{"id":"https://openalex.org/keywords/synthetic-data","display_name":"Synthetic data","score":0.49239999055862427},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.47850000858306885},{"id":"https://openalex.org/keywords/stochastic-gradient-descent","display_name":"Stochastic gradient descent","score":0.4178999960422516},{"id":"https://openalex.org/keywords/statistical-model","display_name":"Statistical model","score":0.41350001096725464}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7063999772071838},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.6729999780654907},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.6650000214576721},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.600600004196167},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.5748000144958496},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5102999806404114},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.49239999055862427},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.47850000858306885},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.4178999960422516},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.41350001096725464},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.40450000762939453},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.396699994802475},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3598000109195709},{"id":"https://openalex.org/C9417928","wikidata":"https://www.wikidata.org/wiki/Q1070689","display_name":"Image processing","level":3,"score":0.35040000081062317},{"id":"https://openalex.org/C35772409","wikidata":"https://www.wikidata.org/wiki/Q1323086","display_name":"Image noise","level":3,"score":0.34380000829696655},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3370000123977661},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.3352999985218048},{"id":"https://openalex.org/C113364801","wikidata":"https://www.wikidata.org/wiki/Q26674","display_name":"High fidelity","level":2,"score":0.33480000495910645},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3181000053882599},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.30480000376701355},{"id":"https://openalex.org/C198082294","wikidata":"https://www.wikidata.org/wiki/Q3399648","display_name":"Position (finance)","level":2,"score":0.28209999203681946},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2818000018596649},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.2761000096797943}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.30312","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.30312","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.2605.30312","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.30312","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":[{"score":0.5392163395881653,"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Differentially":[0],"private":[1,117],"(DP)":[2],"image":[3,206],"synthesis":[4],"generates":[5],"images":[6,42,54,169,218],"that":[7,94,147,154,209],"preserve":[8],"the":[9,86,145,161,167,179,187,212],"statistical":[10],"characteristics":[11],"of":[12,26,67,81,88,216],"a":[13,64,105,128],"sensitive":[14,17,41,53,168,197,205],"dataset,":[15],"enabling":[16],"data":[18],"analysis":[19],"and":[20,114,175,214],"usage":[21],"while":[22,219],"providing":[23],"rigorous":[24],"guarantees":[25],"privacy":[27],"leakage.":[28],"Existing":[29],"methods":[30,226],"fine-tune":[31],"public":[32,50,60,82,171],"models":[33,51,61,83],"using":[34,73],"DP":[35,106,162,200],"Stochastic":[36],"Gradient":[37],"Descent":[38],"(DP-SGD)":[39],"on":[40,52,77,196,203],"to":[43,84,111,131,178,181],"generate":[44],"synthetic":[45,217],"images.":[46],"But":[47],"full":[48],"fine-tuning":[49,225],"is":[55,102,142],"computationally":[56],"expensive,":[57],"because":[58],"current":[59],"typically":[62],"contain":[63],"large":[65],"number":[66,87],"parameters.":[68,90],"Recent":[69],"work":[70],"proposes":[71],"heuristically":[72],"Low-Rank":[74],"Adaptation":[75],"(LoRA)":[76],"all":[78,99],"attention-layer":[79,100],"parameters":[80,101,135,156],"reduce":[85],"trainable":[89],"However,":[91],"we":[92,123,165],"argue":[93],"exhaustive":[95],"LoRA":[96,137],"coverage":[97],"across":[98],"suboptimal":[103],"in":[104],"setting,":[107],"as":[108],"it":[109],"leads":[110],"noise":[112,177],"accumulation":[113],"collapse":[115],"during":[116],"training.":[118],"To":[119],"address":[120],"this":[121],"issue,":[122],"propose":[124],"DP-SAPF,":[125],"which":[126],"uses":[127],"saliency-aware":[129],"strategy":[130],"identify":[132],"specific":[133],"target":[134],"for":[136,160,199],"training":[138],"under":[139],"DP.":[140,183],"DP-SAPF":[141,185,210],"inspired":[143],"by":[144],"fact":[146],"larger":[148],"gradients":[149,180],"signify":[150],"higher":[151],"saliency,":[152],"indicating":[153],"these":[155],"are":[157],"most":[158,188],"critical":[159],"learning.":[163],"Specifically,":[164],"feed":[166],"into":[170],"models,":[172],"compute":[173],"gradients,":[174],"add":[176],"satisfy":[182],"Then,":[184],"identifies":[186],"salient":[189],"parameters,":[190],"those":[191],"exhibiting":[192],"high":[193],"gradient":[194],"magnitudes":[195],"images,":[198],"fine-tuning.":[201],"Experiments":[202],"four":[204],"datasets":[207],"show":[208],"improves":[211],"utility":[213],"fidelity":[215],"requiring":[220],"fewer":[221],"computational":[222],"resources":[223],"than":[224],"without":[227],"parameter":[228],"selection.":[229]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-30T00:00:00"}