{"id":"W3189787503","doi":"10.1109/jsait.2021.3103494","title":"Compressing Gradients by Exploiting Temporal Correlation in Momentum-SGD","year":2021,"lang":"en","type":"article","venue":"IEEE Journal on Selected Areas in Information Theory","topic":"Stochastic Gradient Optimization Techniques","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"Natural Sciences and Engineering Research Council of Canada; Huawei Technologies","keywords":"Bottleneck; Computer science; Computation; Rate of convergence; Convergence (economics); Algorithm; Momentum (technical analysis); Mathematical optimization; Information bottleneck method; Compression (physics); Exploit; Norm (philosophy); Mathematics; Artificial intelligence; Telecommunications","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":[0.001681722,0.0006918095,0.0009511677,0.0004983591,0.0004228855,0.00107048,0.000748809,0.0008481907,0.001256608],"category_scores_gemma":[0.006766446,0.000396731,0.0003376199,0.0005767148,0.001361651,0.001390099,0.001648143,0.001267453,0.0002928617],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007706904,"about_ca_system_score_gemma":0.00158836,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002119549,"about_ca_topic_score_gemma":0.002638829,"domain_scores_codex":[0.99949,0.0001295234,0.00004297039,0.00007059261,0.0002216499,0.00004523894],"domain_scores_gemma":[0.9980348,0.001171079,0.0001567548,0.0002741505,0.0002776983,0.00008549196],"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":[0.0001996283,0.00006598623,0.001427117,0.0001444337,0.00003443454,0.0001466085,0.0001569426,0.8209412,0.009338035,0.03643875,0.001824458,0.1292824],"study_design_scores_gemma":[0.0000071512,0.00002508526,0.00005307389,0.000005464436,0.000002117541,0.00001759845,0.000004897513,0.9942592,0.00144606,0.003823779,0.0003521382,0.000003322333],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01578643,0.0002084368,0.982245,0.0002276453,0.00004070148,0.00004439259,0.00003570248,0.0003350285,0.00107658],"genre_scores_gemma":[0.6064258,0.0004290037,0.3893241,0.0002813263,0.00009193294,0.0002030376,0.0002009797,0.0001688886,0.002874971],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.002119549,"threshold_uncertainty_score":0.008893847,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.009529290614629546,"score_gpt":0.2318131972925966,"score_spread":0.222283906677967,"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."}}