{"id":"W3126140139","doi":"","title":"Uncertainty for deep image classifiers on out of distribution data.","year":2021,"lang":"en","type":"article","venue":"","topic":"Anomaly Detection Techniques and Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Outlier; Calibration; Computer science; Benchmark (surveying); Range (aeronautics); Artificial intelligence; Data mining; Probability distribution; Machine learning; Statistics; Pattern recognition (psychology); Mathematics; Geography","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":[0.005068581,0.001068712,0.001407268,0.001926865,0.0008243219,0.002021222,0.002600611,0.002349769,0.001744567],"category_scores_gemma":[0.02849503,0.0009016319,0.001110437,0.001216141,0.001881631,0.004216877,0.003988884,0.004475738,0.0006349555],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002137069,"about_ca_system_score_gemma":0.001117605,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00444902,"about_ca_topic_score_gemma":0.005086522,"domain_scores_codex":[0.9973385,0.0006878061,0.0001628238,0.0005587946,0.001050641,0.0002013589],"domain_scores_gemma":[0.9889739,0.007101002,0.001044873,0.00130913,0.001271877,0.0002992123],"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":[0.0003346339,0.0001166318,0.005597483,0.0001813416,0.0001825729,0.0001631929,0.0002715066,0.6651759,0.005264754,0.03319206,0.004922427,0.2845975],"study_design_scores_gemma":[0.000003939806,0.00002541425,0.0004455074,0.00001981781,0.000008141875,0.00004216739,0.00001875134,0.9692394,0.001889416,0.02765489,0.0006429688,0.000009655576],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02405073,0.0006471477,0.9724855,0.0005101111,0.0000562519,0.00004932181,0.0002208208,0.0009453418,0.001034673],"genre_scores_gemma":[0.7401311,0.0007097644,0.2524947,0.0007207416,0.0003148474,0.0002208904,0.00157304,0.0004345749,0.003400445],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005068581,"threshold_uncertainty_score":0.02680558,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04642992248722309,"score_gpt":0.317586111422284,"score_spread":0.2711561889350609,"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."}}