{"id":"W2788204593","doi":"10.1145/3152494.3152498","title":"Are saddles good enough for neural networks","year":2018,"lang":"en","type":"article","venue":"","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canadian Institute for Advanced Research","keywords":"Maxima and minima; MNIST database; Degeneracy (biology); Artificial neural network; Saddle point; Gradient descent; Computer science; Saddle; Convergence (economics); Perspective (graphical); Work (physics); Deep neural networks; Artificial intelligence; Mathematical optimization; Machine learning; Mathematics; Engineering; Geometry; Economics; Mathematical analysis","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":[0.004145815,0.0007948201,0.00166312,0.001495854,0.00135037,0.002524894,0.001159744,0.002992569,0.007333674],"category_scores_gemma":[0.02646209,0.001124339,0.001238059,0.0006387225,0.003406492,0.0061903,0.002052194,0.002617489,0.001421399],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001014752,"about_ca_system_score_gemma":0.0008223394,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001495734,"about_ca_topic_score_gemma":0.001541446,"domain_scores_codex":[0.9986571,0.0005338478,0.00008637695,0.0003331871,0.0002418644,0.0001475089],"domain_scores_gemma":[0.9937347,0.003997203,0.0006165976,0.0007117139,0.0005450876,0.0003948528],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001434478,0.00005690275,0.003933923,0.0003489756,0.0001860826,0.0002255648,0.000459355,0.1645408,0.002387046,0.7939987,0.01107219,0.02264701],"study_design_scores_gemma":[0.0000261564,0.00004410554,0.0006470584,0.00008453195,0.00001675631,0.00007839684,0.00008726088,0.2192024,0.0004361503,0.7771469,0.002206188,0.00002418336],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1417938,0.003632667,0.819106,0.009828605,0.0003067941,0.00009694542,0.0006333345,0.0008007561,0.02380108],"genre_scores_gemma":[0.9227573,0.00186989,0.06582504,0.001608336,0.0001457109,0.0002515604,0.0006455631,0.0005206743,0.006376026],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007333674,"threshold_uncertainty_score":0.02453357,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02769192627769521,"score_gpt":0.2642360515460286,"score_spread":0.2365441252683334,"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."}}