{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003232869,0.0008441234,0.0008924195,0.0007425956,0.0006643751,0.001388196,0.001334411,0.001377437,0.002683848],"category_scores_gemma":[0.02162535,0.0006614673,0.0005271741,0.0006522454,0.00234735,0.002939462,0.0022496,0.00198403,0.0006615108],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001615747,"about_ca_system_score_gemma":0.001224515,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00298255,"about_ca_topic_score_gemma":0.002603734,"domain_scores_codex":[0.9987826,0.0005763404,0.000054433,0.0001904857,0.0002904839,0.0001057163],"domain_scores_gemma":[0.9932021,0.004025636,0.0009280723,0.0007015222,0.0006893083,0.00045329],"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":[0.0001092571,0.00009600214,0.003047397,0.00008869235,0.00005442158,0.0001254466,0.0002774462,0.8041027,0.004197282,0.1640414,0.001332465,0.02252749],"study_design_scores_gemma":[0.000002751638,0.00001474485,0.0002403203,0.000004955703,0.000001232909,0.000009502159,0.000009050212,0.9722996,0.0002514386,0.02703056,0.0001311676,0.000004611213],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07552068,0.0002506466,0.9201126,0.0005291489,0.00003975681,0.00003509373,0.00005516546,0.0003239228,0.003133035],"genre_scores_gemma":[0.8955081,0.0003081003,0.09939162,0.0002123257,0.00007567018,0.0001353304,0.0002028383,0.0002542307,0.003911789],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003232869,"threshold_uncertainty_score":0.01709723,"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."}}