{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0009778908,0.0003846679,0.0005411673,0.0008627339,0.0005016687,0.0002498741,0.0009636983,0.0001496263,0.00002269653],"category_scores_gemma":[0.00004354432,0.0003489323,0.0001543233,0.001633087,0.00009148289,0.0006684091,0.00009704676,0.0006969268,0.00002616517],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006124274,"about_ca_system_score_gemma":0.00003565578,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001350847,"about_ca_topic_score_gemma":0.000936322,"domain_scores_codex":[0.9967313,0.0004897115,0.0005854619,0.001275527,0.0004761147,0.0004418806],"domain_scores_gemma":[0.9976271,0.0005083868,0.0002101155,0.001327017,0.0001072583,0.0002200579],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0000435578,0.00005950916,0.002565607,0.0001554338,0.0002607632,0.000006135679,0.0008305846,0.3794734,0.0002357621,0.00004200722,0.000004065383,0.6163231],"study_design_scores_gemma":[0.0002901016,0.0003690577,0.004291644,0.00006549917,0.0002798711,0.000005208504,0.00003490174,0.9883966,0.005696225,0.000219487,0.00002267617,0.0003286934],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.08275466,0.00004557813,0.9155743,0.0003862877,0.000522137,0.0003818444,0.00002661197,0.0002714717,0.00003716243],"genre_scores_gemma":[0.9958802,0.000692827,0.002685759,0.0003486238,0.0000394579,0.00005104592,0.00009378279,0.00003004727,0.0001782746],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9131255,"threshold_uncertainty_score":0.9998963,"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."}}