{"id":"W4366813870","doi":"10.1016/j.neunet.2023.04.028","title":"Stability analysis of stochastic gradient descent for homogeneous neural networks and linear classifiers","year":2023,"lang":"en","type":"article","venue":"Neural Networks","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":9,"is_retracted":false,"has_abstract":false,"ca_institutions":"Université Laval","funders":"Fonds de recherche du Québec – Nature et technologies","keywords":"Stability (learning theory); Generalization; Stochastic gradient descent; Artificial neural network; Mathematics; Activation function; Invariant (physics); Regular polygon; Computer science; Euclidean geometry; Gradient descent; Invariant measure; Applied mathematics; Artificial intelligence; Machine learning; Mathematical analysis","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004970344,0.001147458,0.001522812,0.001637529,0.000997696,0.001974107,0.001652067,0.0016289,0.003364896],"category_scores_gemma":[0.02055382,0.0007947629,0.001006089,0.0007639672,0.002744367,0.002561984,0.002328927,0.001906767,0.0005609695],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00260773,"about_ca_system_score_gemma":0.00180129,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008012774,"about_ca_topic_score_gemma":0.004352812,"domain_scores_codex":[0.9987112,0.0006235447,0.00005386614,0.0002376287,0.0002746488,0.00009913832],"domain_scores_gemma":[0.9915052,0.00572578,0.0006694127,0.0002655876,0.001556448,0.000277503],"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":[0.000188855,0.00006264971,0.001038972,0.0002135705,0.0001369449,0.0001159068,0.0002370634,0.548366,0.003646446,0.4251704,0.003215506,0.01760775],"study_design_scores_gemma":[0.0000027119,0.000008908664,0.0000922198,0.000006393952,0.000005141513,0.000006379742,0.000006802725,0.9695295,0.0001726766,0.03001203,0.0001522868,0.000004985474],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02540202,0.0007739334,0.9667422,0.0008881275,0.00008295791,0.0000612871,0.00009000943,0.0001461721,0.005813231],"genre_scores_gemma":[0.8783811,0.001446159,0.09208282,0.0003850755,0.000362933,0.0004314217,0.0004082426,0.0005218419,0.02598044],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.008012774,"threshold_uncertainty_score":0.02628601,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03546809893951639,"score_gpt":0.2656003313902598,"score_spread":0.2301322324507434,"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."}}