{"id":"W4387390352","doi":"10.48550/arxiv.2310.03010","title":"Spectral alignment of stochastic gradient descent for high-dimensional classification tasks","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Face and Expression Recognition","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs; National Science Foundation","keywords":"Hessian matrix; Outlier; Rank (graph theory); Stochastic gradient descent; Eigenvalues and eigenvectors; Gradient descent; Artificial intelligence; Pattern recognition (psychology); Computer science; Mathematics; Artificial neural network; Algorithm; Applied mathematics; Combinatorics; Physics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.000205677,0.0002125736,0.0002580709,0.0002490696,0.0001068539,0.00003568878,0.0007144098,0.0001763278,0.00001350769],"category_scores_gemma":[0.00002563727,0.0002308253,0.0001837605,0.0002588665,0.00006443436,0.0001414409,0.000640726,0.0001951185,0.00006135828],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002107863,"about_ca_system_score_gemma":0.0001326459,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008868997,"about_ca_topic_score_gemma":0.00001597911,"domain_scores_codex":[0.9984494,0.00006250413,0.0002476167,0.0008200845,0.0001438344,0.0002765998],"domain_scores_gemma":[0.9986261,0.0001194544,0.0003072048,0.0006472033,0.0001793279,0.0001206898],"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":[0.0001669579,0.0005086602,0.0002299591,0.0002712873,0.00020007,0.00004875294,0.0003154353,0.6324722,0.0039736,0.355119,0.005234272,0.001459819],"study_design_scores_gemma":[0.000786556,0.0001547553,0.003505013,0.0003018846,0.00009154087,0.000001577161,0.00006839057,0.8558373,0.003346968,0.1354951,0.00003647316,0.0003745362],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2796256,0.00001040162,0.718586,0.0002220072,0.0008802055,0.000465583,0.00006222443,0.0001198684,0.00002808777],"genre_scores_gemma":[0.9950433,0.00002058137,0.004215987,0.00003581341,0.00005322775,0.000008045185,0.0001323712,0.00001493369,0.000475789],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7154176,"threshold_uncertainty_score":0.9412782,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1224365376044786,"score_gpt":0.2111763213144974,"score_spread":0.08873978371001884,"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."}}