{"id":"W4315784658","doi":"10.56553/popets-2023-0010","title":"Individualized PATE: Differentially Private Machine Learning with Individual Privacy Guarantees","year":2023,"lang":"en","type":"article","venue":"Proceedings on Privacy Enhancing Technologies","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Toronto","funders":"Bundesministerium für Bildung und Forschung","keywords":"Differential privacy; Computer science; Information privacy; Private information retrieval; MNIST database; Machine learning; Privacy by Design; Process (computing); Information sensitivity; Privacy protection; Data mining; Artificial intelligence; Computer security; Internet privacy; Deep learning","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","metaepi_narrow","scholarly_communication","open_science","research_integrity"],"consensus_categories":["open_science"],"category_scores_codex":[0.00186689,0.001055505,0.0009783639,0.002269911,0.001054512,0.001369041,0.04767339,0.000718312,0.00001526638],"category_scores_gemma":[0.0499954,0.0008588752,0.0001710539,0.005308542,0.0005634475,0.002068843,0.1171895,0.002746767,0.0002489213],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002164928,"about_ca_system_score_gemma":0.0001223978,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002087408,"about_ca_topic_score_gemma":0.00000548404,"domain_scores_codex":[0.9928599,0.00007302274,0.001004258,0.002289781,0.001807308,0.00196572],"domain_scores_gemma":[0.9924434,0.0005756562,0.0008694065,0.005688349,0.0002975308,0.0001256999],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0006560942,0.001194437,0.112791,0.001586664,0.001727244,0.0006075288,0.007729519,0.0000955964,0.1230159,0.1182584,0.06793308,0.5644045],"study_design_scores_gemma":[0.003087478,0.00199479,0.008507978,0.00135359,0.0001106599,0.0001449754,0.001281621,0.02565428,0.5945019,0.3442096,0.01656373,0.002589481],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8729888,0.0002540023,0.04274091,0.02516901,0.0003157609,0.001123361,0.00004110365,0.05689241,0.0004746811],"genre_scores_gemma":[0.798342,0.0006704695,0.2002054,0.0001170755,0.00003706992,0.0003581592,0.00006474175,0.0001195263,0.00008558978],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5618151,"threshold_uncertainty_score":0.9996676,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02652680026550532,"score_gpt":0.2602872961785198,"score_spread":0.2337604959130145,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}