{"id":"W3212182603","doi":"","title":"Heavy Tails in SGD and Compressibility of Overparametrized Neural Networks.","year":2021,"lang":"en","type":"article","venue":"HAL (Le Centre pour la Communication Scientifique Directe)","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Pruning; Artificial neural network; Computer science; Generalization; Compressibility; Limit (mathematics); Compression (physics); Node (physics); Computation; Mathematics; Algorithm; Mathematical optimization; Applied mathematics; Artificial intelligence; Mathematical analysis; Physics","routes":{"ca_aff":true,"ca_fund":false,"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":[0.002648729,0.0007960583,0.0011547,0.001171831,0.0005813431,0.001724631,0.001441626,0.001757458,0.003919196],"category_scores_gemma":[0.02986435,0.0007002051,0.0005772747,0.0007139586,0.003160536,0.003620195,0.002363245,0.002727325,0.0004706481],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00173452,"about_ca_system_score_gemma":0.0008603098,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003190822,"about_ca_topic_score_gemma":0.003570553,"domain_scores_codex":[0.9993649,0.0002947569,0.0000322994,0.0001121005,0.0001175019,0.00007846056],"domain_scores_gemma":[0.9863236,0.01002754,0.001074372,0.0009865628,0.0006995117,0.0008885215],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0001759861,0.00004440702,0.0033446,0.0002378017,0.00007941474,0.0004650841,0.000287371,0.4254009,0.002301985,0.5452129,0.005710145,0.0167394],"study_design_scores_gemma":[0.000009795142,0.00001750408,0.0006345503,0.00002886783,0.000006350303,0.00004678356,0.00001738274,0.8009158,0.0003545117,0.1974083,0.0005449689,0.00001520673],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2704899,0.003805839,0.7013801,0.005813014,0.000324919,0.00008254383,0.0008554594,0.001002118,0.01624607],"genre_scores_gemma":[0.962423,0.001539984,0.02026416,0.0005736942,0.0002855306,0.0001131393,0.0006338055,0.0003696178,0.01379709],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003919196,"threshold_uncertainty_score":0.01400799,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01208512051257982,"score_gpt":0.226041951742967,"score_spread":0.2139568312303872,"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."}}