{"id":"W3034995656","doi":"10.24963/ijcai.2020/452","title":"Closing the Generalization Gap of Adaptive Gradient Methods in Training Deep Neural Networks","year":2020,"lang":"en","type":"article","venue":"","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":49,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Canadian Institute for Advanced Research; University of Pennsylvania","keywords":"Stochastic gradient descent; Gradient descent; Computer science; Artificial neural network; Convergence (economics); Generalization; Closing (real estate); Artificial intelligence; Momentum (technical analysis); Gradient method; Stationary point; Deep learning; Rate of convergence; Algorithm; Mathematical optimization; Mathematics; Key (lock); Law","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":[0.0003937795,0.00009981739,0.0001501942,0.00007403379,0.00005366315,0.00004224177,0.0004875549,0.00003939924,0.000008393087],"category_scores_gemma":[0.0001122782,0.00007416668,0.00004509626,0.0008933498,0.0000475394,0.0002012235,0.000145482,0.0001077502,2.348321e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002433703,"about_ca_system_score_gemma":0.00001414606,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002097626,"about_ca_topic_score_gemma":0.00000621257,"domain_scores_codex":[0.9988931,0.0002352752,0.0003023915,0.0002443737,0.0001435411,0.0001813642],"domain_scores_gemma":[0.9993659,0.0001864254,0.0001269669,0.0002012289,0.00006642377,0.00005301394],"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":[0.000004897947,0.00001090212,0.000119838,0.000002404036,0.000006084641,0.000001165029,0.00686337,0.6990012,0.0002317446,0.2447693,0.00003372915,0.04895534],"study_design_scores_gemma":[0.0001115138,0.00008564616,0.0001797404,0.000006975431,0.000003732308,0.000002767476,0.0001954118,0.9956931,0.0008609448,0.002767085,0.00001260511,0.00008051387],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0002885417,0.0001022791,0.9973274,0.001150827,0.0001267255,0.0002279112,2.138676e-7,0.0001510502,0.0006250878],"genre_scores_gemma":[0.4180883,0.000004112316,0.5812081,0.0006549247,0.00002385149,0.00001166789,0.00000140261,0.000005376681,0.000002277724],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.4177997,"threshold_uncertainty_score":0.302443,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1055841967632464,"score_gpt":0.3277355163136001,"score_spread":0.2221513195503537,"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."}}