{"id":"W4403585525","doi":"10.1093/imaiai/iaae028","title":"Hitting the High-dimensional notes: an ODE for SGD learning dynamics on GLMs and multi-index models","year":2024,"lang":"en","type":"article","venue":"Information and Inference A Journal of the IMA","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Israel Science Foundation; Canadian Institute for Advanced Research","keywords":"Ode; Index (typography); Dynamics (music); Econometrics; Artificial intelligence; Statistical physics; Computer science; Mathematics; Physics; Applied mathematics","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0005365976,0.0000795649,0.00008382168,0.0001224499,0.0002233334,0.0004602893,0.0002926474,0.00003546322,0.000001113685],"category_scores_gemma":[0.0003097953,0.0000436908,0.00003511347,0.0001322372,0.00004605,0.002326856,0.00009709486,0.0002168617,6.440462e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003580231,"about_ca_system_score_gemma":0.00007457229,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000009701519,"about_ca_topic_score_gemma":0.000002159812,"domain_scores_codex":[0.9992843,0.00003345744,0.000312257,0.0000621962,0.0002119869,0.0000957725],"domain_scores_gemma":[0.9990926,0.0003180592,0.0002292472,0.0001111883,0.0002062963,0.00004262246],"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":[0.000027338,0.00001970958,0.0002239494,0.00003647493,0.00002509051,6.196257e-7,0.005048189,0.4095461,0.00003640325,0.3292534,0.00007801184,0.2557047],"study_design_scores_gemma":[0.0001988549,0.0001436397,0.0003671776,0.0001417964,0.000005909438,0.00005145239,0.00010744,0.9851687,0.0001020003,0.01358906,0.00006676791,0.00005718277],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02430227,0.00004489804,0.9731233,0.002114384,0.0002150308,0.0001258783,0.000002771998,0.00003819478,0.00003323584],"genre_scores_gemma":[0.9441758,0.0000223542,0.05523667,0.0005256226,0.00002066064,0.000004196053,0.000001428232,0.000003156856,0.00001006086],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9198736,"threshold_uncertainty_score":0.443858,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02344334007612431,"score_gpt":0.2753627054817372,"score_spread":0.2519193654056129,"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."}}