{"id":"W3092700804","doi":"10.48550/arxiv.2010.09063","title":"Enabling Fast Differentially Private SGD via Just-in-Time Compilation and Vectorization","year":2020,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":18,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Compiler; Vectorization (mathematics); Programming language; Overhead (engineering); Parallel computing; Matching (statistics); Code (set theory); Graph; Theoretical computer science; Set (abstract data type)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002515711,0.001405567,0.0008640049,0.0006383127,0.0008545264,0.002902684,0.002275378,0.001036211,0.01000903],"category_scores_gemma":[0.01142589,0.000846926,0.001367521,0.001025398,0.002401461,0.004802657,0.004497071,0.003355744,0.005404938],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001535082,"about_ca_system_score_gemma":0.00341317,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002216248,"about_ca_topic_score_gemma":0.003328316,"domain_scores_codex":[0.9968727,0.000823922,0.0002345491,0.0006496712,0.0009622378,0.0004569479],"domain_scores_gemma":[0.9933426,0.002018674,0.0002696124,0.003597989,0.0005685714,0.0002025982],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001348473,0.0003654678,0.006323025,0.0006900769,0.0002100509,0.0005441374,0.001028922,0.1761857,0.04793176,0.2860465,0.06927133,0.4100545],"study_design_scores_gemma":[0.0002025439,0.0001270622,0.0005800675,0.00007618872,0.00004867696,0.0002428682,0.0001257249,0.6805938,0.04917798,0.2307132,0.03804353,0.00006820699],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02031865,0.0002820352,0.922529,0.0008260207,0.0002220651,0.00008724986,0.0006531851,0.04794945,0.007132382],"genre_scores_gemma":[0.3943117,0.0004155978,0.5765216,0.0009248638,0.0001803651,0.0003996287,0.002475873,0.01511879,0.009651582],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01000903,"threshold_uncertainty_score":0.03348351,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07078098645669183,"score_gpt":0.2016197265179104,"score_spread":0.1308387400612186,"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."}}