{"id":"W4385825718","doi":"10.48550/arxiv.2308.06239","title":"Private Distribution Learning with Public Data: The View from Sample Compression","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; University of Waterloo","keywords":"Learnability; Differential privacy; Connection (principal bundle); Distribution (mathematics); Sample (material); Upper and lower bounds; Closure (psychology); Sample complexity; Class (philosophy); Concept class; Mathematics; Discrete mathematics; Computer science; Combinatorics; Artificial intelligence; Algorithm; Physics; Mathematical analysis; Political science","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01322122,0.00169036,0.003427862,0.001524933,0.002238565,0.007409805,0.005677082,0.006180234,0.005474202],"category_scores_gemma":[0.0744387,0.001294527,0.002206339,0.003560385,0.009493861,0.02649502,0.01161604,0.01347306,0.001496311],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005905535,"about_ca_system_score_gemma":0.003710112,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001233535,"about_ca_topic_score_gemma":0.0008471148,"domain_scores_codex":[0.9856507,0.006862869,0.0004530424,0.002679171,0.003214541,0.001139627],"domain_scores_gemma":[0.9075363,0.06859735,0.003118939,0.0171373,0.002049306,0.001560823],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0007296001,0.0003336335,0.002010234,0.0003947099,0.0001331355,0.0002532271,0.0006497437,0.08293227,0.001842961,0.8505408,0.006114582,0.05406514],"study_design_scores_gemma":[0.0001140331,0.0000969497,0.0001966146,0.000061266,0.0000250209,0.0001570514,0.00008975818,0.2011347,0.002063346,0.7927459,0.003281878,0.00003338372],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02685614,0.001409463,0.953218,0.01143274,0.0001235258,0.0001592779,0.0007202526,0.0005364502,0.005544252],"genre_scores_gemma":[0.7219188,0.002858573,0.2558111,0.004527315,0.001609685,0.0009940426,0.001569084,0.0005156378,0.01019588],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01322122,"threshold_uncertainty_score":0.06992131,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2065754176131079,"score_gpt":0.2339709936692677,"score_spread":0.02739557605615972,"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."}}