{"id":"W4315750667","doi":"10.1109/tpami.2023.3234291","title":"Adversarially-Regularized Mixed Effects Deep Learning (ARMED) Models Improve Interpretability, Performance, and Generalization on Clustered (non-<i>iid</i>) Data","year":2023,"lang":"en","type":"article","venue":"IEEE Transactions on Pattern Analysis and Machine Intelligence","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"National Institute on Aging; National Institute of Biomedical Imaging and Bioengineering; Canadian Institutes of Health Research; National Institutes of Health; Genentech; IXICO; National Institute of General Medical Sciences; H. Lundbeck A/S; Eisai; Northern California Institute for Research and Education; Servier; University of Texas Southwestern Medical Center; Pfizer; Biogen; BioClinica; Eli Lilly and Company; U.S. Department of Defense; Lyda Hill Foundation; F. Hoffmann-La Roche; University of Southern California; Meso Scale Diagnostics; Alzheimer's Disease Neuroimaging Initiative; Novartis Pharmaceuticals Corporation; Bristol-Myers Squibb; Alzheimer's Association; Foundation for the National Institutes of Health","keywords":"Interpretability; Artificial intelligence; Computer science; Autoencoder; Spurious relationship; Machine learning; Generalization; Classifier (UML); Deep learning; Pattern recognition (psychology); Mathematics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003854042,0.001531789,0.0008862553,0.0003802215,0.0002983797,0.0009655792,0.001448038,0.001064593,0.001298959],"category_scores_gemma":[0.008122549,0.0005521102,0.001291304,0.0003692237,0.0009094445,0.001201704,0.001725527,0.002565523,0.0005755309],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009410447,"about_ca_system_score_gemma":0.0009550792,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00450315,"about_ca_topic_score_gemma":0.007434819,"domain_scores_codex":[0.9988367,0.0006096988,0.00005445433,0.0002770471,0.0001272939,0.00009481428],"domain_scores_gemma":[0.9969591,0.00176923,0.0002971945,0.0005622843,0.0003086398,0.0001035185],"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":[0.0002590276,0.0001424473,0.003765773,0.00008733143,0.0003244985,0.0001081072,0.00008505235,0.8772511,0.004421856,0.008000804,0.003725033,0.1018288],"study_design_scores_gemma":[0.000006941847,0.00004185312,0.0003215128,0.000007820501,0.00001855125,0.00001332102,0.00000439093,0.9924191,0.001148284,0.005677373,0.0003339917,0.00000695948],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06102313,0.0007156743,0.9323679,0.0009808082,0.00009296508,0.00007146291,0.0005074606,0.002403047,0.001837521],"genre_scores_gemma":[0.7449719,0.0005898422,0.2461319,0.001209982,0.0001148575,0.0002056813,0.001514453,0.0003502866,0.004911106],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.00450315,"threshold_uncertainty_score":0.02038234,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02329700720975141,"score_gpt":0.2809728777341108,"score_spread":0.2576758705243594,"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."}}