{"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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001386701,0.0002148388,0.0002371595,0.0002385513,0.0005575525,0.000343239,0.002092936,0.00007623041,0.00004850201],"category_scores_gemma":[0.0004478138,0.000190929,0.00002858073,0.001013791,0.00007659347,0.00114823,0.00142516,0.0007737989,0.0001629912],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001427068,"about_ca_system_score_gemma":0.00006486328,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008095434,"about_ca_topic_score_gemma":0.0000300351,"domain_scores_codex":[0.9976045,0.0002813513,0.0002697555,0.0008801569,0.0004787494,0.0004854983],"domain_scores_gemma":[0.9978097,0.0002592376,0.0002104322,0.001568868,0.00005287277,0.00009887269],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000155178,0.0001982669,0.6480634,0.0003277371,0.000286116,0.0003494786,0.008970428,0.02483997,0.01700132,0.1544479,0.03238554,0.1129746],"study_design_scores_gemma":[0.0004310152,0.0001328173,0.009986797,0.00006458494,0.00000898988,0.00001749442,0.0002283989,0.927022,0.0005316269,0.004402128,0.05677562,0.0003985611],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.002529686,0.00003647565,0.9873849,0.002016819,0.00008950529,0.0001716698,0.00001143223,0.003953662,0.003805894],"genre_scores_gemma":[0.9279655,0.00004551175,0.06925254,0.0001535701,0.00006118517,0.00002137979,0.0006673097,0.00003868547,0.001794316],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9254358,"threshold_uncertainty_score":0.778586,"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."}}