{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.00007857647,0.0002462535,0.0002959433,0.0002829931,0.00009075389,0.0001303975,0.0006031702,0.0002595268,0.0000166728],"category_scores_gemma":[0.00007126689,0.0002737108,0.00006889112,0.000514831,0.0001198904,0.000457131,0.001069551,0.000226768,3.081284e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00008647746,"about_ca_system_score_gemma":0.0001608455,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000583377,"about_ca_topic_score_gemma":0.000004699138,"domain_scores_codex":[0.998641,0.00008938135,0.0002303177,0.0007489258,0.0001189127,0.0001715331],"domain_scores_gemma":[0.9985145,0.00002464026,0.0003320364,0.0006129052,0.0004329367,0.00008301344],"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":[0.000005722275,0.00002940973,0.00008101762,0.00007567367,0.0000383185,0.00001002256,0.000388607,0.9797417,0.0002156529,0.01920627,0.000007344751,0.0002002649],"study_design_scores_gemma":[0.0001979461,0.00003426368,0.00006378626,0.0001634201,0.00004887175,0.000008046411,0.00005233721,0.994044,0.001926907,0.00319866,5.623265e-7,0.0002611944],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.08023774,0.0001767109,0.9187534,0.00002718116,0.0002691704,0.0003116622,0.0000130654,0.0001866292,0.00002440056],"genre_scores_gemma":[0.8920656,0.00006838699,0.107694,0.00002427962,0.00002613648,0.00000121175,0.00005046847,0.00001602321,0.00005386846],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8118279,"threshold_uncertainty_score":0.9999715,"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."}}