{"id":"W3102590949","doi":"10.1088/1742-5468/ac3ae7","title":"Hausdorff dimension, heavy tails, and generalization in neural networks*","year":2021,"lang":"en","type":"article","venue":"Journal of Statistical Mechanics Theory and Experiment","topic":"Markov Chains and Monte Carlo Methods","field":"Mathematics","cited_by":20,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Toronto","funders":"Agence Nationale de la Recherche","keywords":"Generalization; Stochastic gradient descent; Applied mathematics; Artificial neural network; Hausdorff dimension; Mathematics; Hausdorff distance; Computer science; Range (aeronautics); Dimension (graph theory); Metric (unit); Artificial intelligence; Pure mathematics; Mathematical analysis","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":[0.001684158,0.0001252615,0.0003310634,0.00005297492,0.00005829264,0.00004102412,0.00004328575,0.00007065365,0.00004287466],"category_scores_gemma":[0.0007512445,0.00010038,0.00003428426,0.00007432183,0.00002889478,0.00007158715,0.00007335008,0.0001778087,1.22512e-8],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003389141,"about_ca_system_score_gemma":0.0000253321,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000001483577,"about_ca_topic_score_gemma":0.000003078332,"domain_scores_codex":[0.9984051,0.0006334152,0.0004765764,0.0001471935,0.0001679946,0.0001697451],"domain_scores_gemma":[0.998631,0.0008432939,0.0001515812,0.0001031927,0.0001182001,0.0001527363],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0003002266,0.0001232978,0.000009970628,0.00003640206,0.00002753769,0.0002045008,0.0006219642,0.00007515041,0.005481937,0.982848,0.0002789628,0.009992013],"study_design_scores_gemma":[0.001829725,0.0005539532,0.00002276326,0.0001737525,0.0000797547,0.0005379042,0.002418272,0.1222339,0.01740939,0.8537735,0.0006788162,0.0002882349],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1386532,0.003922215,0.8568228,0.0001617313,0.0003114748,0.00007654077,0.000004368147,0.000004876426,0.00004274146],"genre_scores_gemma":[0.866667,0.0008015991,0.1319662,0.0003458942,0.0001262953,0.00000449264,0.000003306126,0.00002014564,0.00006502702],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7280138,"threshold_uncertainty_score":0.4093379,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02945931311743106,"score_gpt":0.3379642316986228,"score_spread":0.3085049185811917,"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."}}