{"id":"W3168222560","doi":"10.48550/arxiv.2106.04881","title":"Fractal Structure and Generalization Properties of Stochastic Optimization Algorithms","year":2021,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Ergodic theory; Generalization; Algorithm; Stochastic gradient descent; Dynamical systems theory; Mathematics; Computer science; Hessian matrix; Fractal; Artificial neural network; Bounded function; Invariant measure; Hyperparameter; Mathematical optimization; Applied mathematics; Artificial intelligence","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":[0.004640472,0.0007698107,0.0009291158,0.001585193,0.0007214517,0.001611733,0.0009515414,0.001542516,0.001349511],"category_scores_gemma":[0.02544168,0.0005356822,0.001076627,0.0006393815,0.00295337,0.002571213,0.002251923,0.002487603,0.0002378001],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00174962,"about_ca_system_score_gemma":0.0008611911,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001840376,"about_ca_topic_score_gemma":0.001180195,"domain_scores_codex":[0.9983321,0.0005456288,0.0001287799,0.0002945217,0.0005377505,0.0001611795],"domain_scores_gemma":[0.9836606,0.01160521,0.001550858,0.001354515,0.00141405,0.0004146115],"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":[0.00005396609,0.00003667569,0.002186256,0.0001259766,0.00007561923,0.0001129184,0.0002001228,0.6155334,0.003136811,0.3586608,0.0009343737,0.01894302],"study_design_scores_gemma":[0.000004355628,0.00002051402,0.0003271373,0.00001586296,0.000006182938,0.00002471414,0.000007394956,0.9344043,0.0003994244,0.06446875,0.0003131577,0.000008151833],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.07867648,0.001154495,0.9131377,0.001013445,0.00007329907,0.00004948325,0.00006667303,0.0002580637,0.005570401],"genre_scores_gemma":[0.8889048,0.001316316,0.1059707,0.0003047535,0.0002062789,0.000208829,0.0002481921,0.0002328871,0.002607074],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.004640472,"threshold_uncertainty_score":0.0245415,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04562625423809228,"score_gpt":0.1765352584686625,"score_spread":0.1309090042305703,"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."}}