{"id":"W4364381349","doi":"10.1007/s10994-023-06326-9","title":"Understanding CNN fragility when learning with imbalanced data","year":2023,"lang":"en","type":"article","venue":"Machine Learning","topic":"Imbalanced Data Classification Techniques","field":"Computer Science","cited_by":65,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"Office of Naval Research","keywords":"Convolutional neural network; Class (philosophy); Artificial intelligence; Computer science; Feature (linguistics); Focus (optics); Pattern recognition (psychology); Set (abstract data type); Machine learning; Image (mathematics)","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":[0.003283059,0.000873264,0.000751846,0.001199496,0.000693358,0.001471426,0.001490161,0.001191693,0.001170216],"category_scores_gemma":[0.02351966,0.0005428807,0.0005181822,0.0007663618,0.001450241,0.003401618,0.001670321,0.002316069,0.0002478444],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002731525,"about_ca_system_score_gemma":0.0005760792,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00520066,"about_ca_topic_score_gemma":0.004706862,"domain_scores_codex":[0.9989656,0.0002105508,0.00006708349,0.0002872104,0.0002753344,0.0001942873],"domain_scores_gemma":[0.9914335,0.004424834,0.001466011,0.0013237,0.0009640331,0.0003879259],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.000880408,0.0003000212,0.07375818,0.000191488,0.0001773281,0.0006418719,0.0004112359,0.783715,0.01658257,0.01215016,0.005534855,0.1056568],"study_design_scores_gemma":[0.00001092533,0.00005336852,0.007082976,0.00001792384,0.0000133369,0.00006769493,0.00007252708,0.9751363,0.0040936,0.01301996,0.0004200174,0.00001141275],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8909622,0.0007332864,0.102526,0.002031849,0.0001213847,0.00007999766,0.0006027473,0.0008565361,0.002086021],"genre_scores_gemma":[0.9913455,0.00008705605,0.00741567,0.0001935584,0.00003515427,0.00004075867,0.0004150901,0.00004561482,0.0004217081],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.00520066,"threshold_uncertainty_score":0.01981872,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1495274344458986,"score_gpt":0.3026147751511101,"score_spread":0.1530873407052114,"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."}}