{"id":"W2990091474","doi":"10.48550/arxiv.1911.11195","title":"A Novel Unsupervised Post-Processing Calibration Method for DNNS with Robustness to Domain Shift","year":2019,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université Laval","funders":"","keywords":"Robustness (evolution); Computer science; Calibration; Artificial intelligence; Range (aeronautics); Scaling; Pattern recognition (psychology); Machine learning; Statistics; Mathematics; Engineering","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0007459407,0.0005476273,0.0006025088,0.0004716136,0.0003574743,0.0003986346,0.002527326,0.0004058478,0.00001268112],"category_scores_gemma":[0.0001220957,0.0005704581,0.0002076479,0.001015662,0.00006235592,0.001031139,0.002083111,0.0007524038,0.00001137767],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003437673,"about_ca_system_score_gemma":0.000792377,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003080351,"about_ca_topic_score_gemma":0.000118105,"domain_scores_codex":[0.9966956,0.0002271586,0.0003069862,0.001906337,0.0002105765,0.0006533605],"domain_scores_gemma":[0.9973228,0.0003509634,0.0003946072,0.00123685,0.0004220755,0.0002726999],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002050936,0.00006396396,0.0003358117,0.0002421228,0.00006468946,0.00003797702,0.001099936,0.9309793,0.0002242924,0.06597177,0.0000368946,0.0007381327],"study_design_scores_gemma":[0.001423421,0.0001646445,0.0003296705,0.0002493838,0.00009582462,0.000006384565,0.0003157205,0.9922213,0.00009032463,0.0041444,0.0001723832,0.0007865582],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02293172,0.00001404993,0.973593,0.001067557,0.0005369276,0.001232251,0.00003538207,0.0003884909,0.0002006274],"genre_scores_gemma":[0.5391994,0.000001378891,0.4600265,0.0003049715,0.0001070775,0.000006294885,0.00004831313,0.00004706922,0.000259036],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.5162676,"threshold_uncertainty_score":0.9996747,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05045522398792803,"score_gpt":0.2274240020226364,"score_spread":0.1769687780347083,"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."}}