{"id":"W4408145900","doi":"10.1109/icmla61862.2024.00275","title":"Clustered Federated Learning with Non-IID Data: Mitigating Accuracy Overestimates Through Hold-Out Model Selection and Evaluation","year":2024,"lang":"en","type":"article","venue":"","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Western University","funders":"","keywords":"Computer science; Selection (genetic algorithm); Data modeling; Artificial intelligence; Machine learning; Data mining; Model selection; Database","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":[0.02203609,0.001596546,0.001951921,0.001309638,0.001381507,0.002867176,0.003512084,0.001886598,0.0008104105],"category_scores_gemma":[0.05723119,0.0005749816,0.0008666723,0.001500751,0.001970843,0.004227314,0.003777491,0.002854079,0.0004849531],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002579174,"about_ca_system_score_gemma":0.002746432,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004160341,"about_ca_topic_score_gemma":0.004042084,"domain_scores_codex":[0.983851,0.008646479,0.0007367992,0.002549611,0.00339291,0.0008232161],"domain_scores_gemma":[0.955658,0.01786756,0.001878269,0.01764872,0.005835252,0.001112287],"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":[0.002663883,0.001015609,0.03314652,0.0002040955,0.0005696028,0.0003830855,0.0004278524,0.6917159,0.00604182,0.008271194,0.0080839,0.2474766],"study_design_scores_gemma":[0.00004947523,0.0002302881,0.001358834,0.00002294959,0.00003360684,0.00009875283,0.00007316716,0.9839187,0.005563032,0.008031596,0.0006003848,0.00001921904],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3222532,0.001243479,0.664524,0.001278347,0.0002062865,0.0002742633,0.0004416397,0.006477537,0.003301276],"genre_scores_gemma":[0.917317,0.0000766969,0.08067074,0.0003254131,0.00004054862,0.0001039567,0.000533835,0.0001686046,0.0007631753],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02203609,"threshold_uncertainty_score":0.1165394,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09604347398178321,"score_gpt":0.358160383640574,"score_spread":0.2621169096587908,"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."}}