{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0004363978,0.0002673478,0.0005051449,0.0004016463,0.0001898043,0.00008125247,0.0005960834,0.0001468486,0.000005000416],"category_scores_gemma":[0.0001345328,0.0002523292,0.000260366,0.00286983,0.0001795219,0.0001668729,0.0003138159,0.0002116337,3.170298e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005853271,"about_ca_system_score_gemma":0.00001701095,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001672059,"about_ca_topic_score_gemma":0.00002317389,"domain_scores_codex":[0.9977575,0.0001010918,0.0005458687,0.0006983266,0.0002881256,0.0006091039],"domain_scores_gemma":[0.9980779,0.0005717714,0.0002466301,0.0006820527,0.0002074526,0.0002141581],"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":[0.00003378065,0.00003704889,0.001729545,0.00001188313,0.0001306527,0.000004003874,0.0001208898,0.9880201,0.00001025578,0.001644864,0.0002158147,0.008041144],"study_design_scores_gemma":[0.0002655754,0.000248309,0.003195691,0.00001004776,0.0002279085,0.000005370923,0.00002059658,0.9955066,0.00002083635,0.0002732188,0.000005694914,0.000220167],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.08255788,0.000188083,0.9153408,0.0001814905,0.0005749686,0.0006712409,0.0000151397,0.0004650702,0.00000529517],"genre_scores_gemma":[0.9910597,0.00002732736,0.008490523,0.0001320822,0.00009067552,0.0001063775,0.00005787778,0.00002318957,0.0000122794],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9085018,"threshold_uncertainty_score":0.9999929,"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."}}