{"id":"W3158909043","doi":"10.48550/arxiv.1908.06077","title":"NUQSGD: Provably Communication-efficient Data-parallel SGD via Nonuniform Quantization","year":2019,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Stochastic gradient descent; Quantization (signal processing); Heuristic; Deep neural networks; Gradient descent; Communication complexity; Algorithm; Theoretical computer science; Artificial neural network; Artificial intelligence","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":[0.002077146,0.001610209,0.001839137,0.0007382164,0.00080018,0.001621266,0.002763713,0.001697522,0.004159261],"category_scores_gemma":[0.01183842,0.0006807077,0.0007401874,0.001184794,0.001653021,0.002705237,0.004193131,0.003439052,0.001607873],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001512305,"about_ca_system_score_gemma":0.002798057,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004859179,"about_ca_topic_score_gemma":0.008203546,"domain_scores_codex":[0.9984536,0.0005255464,0.0001248295,0.0002809648,0.0004829435,0.0001321482],"domain_scores_gemma":[0.9965472,0.001733871,0.0001613116,0.0009864875,0.0003966597,0.0001744467],"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":[0.000521616,0.0001588119,0.001841199,0.0002719747,0.00007485918,0.0001736774,0.0002792607,0.6830636,0.004610565,0.0687485,0.02052796,0.2197279],"study_design_scores_gemma":[0.00005597312,0.00003544565,0.00007840813,0.00001414315,0.000004747042,0.00002842044,0.00001979674,0.9641618,0.001161414,0.0330453,0.001387464,0.000007137975],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01535195,0.0005245092,0.9762467,0.0006810849,0.0001531504,0.0001168013,0.0003691614,0.003421184,0.003135426],"genre_scores_gemma":[0.408545,0.0004322341,0.5818713,0.0008730594,0.0001997671,0.0004711186,0.001917048,0.001006717,0.00468381],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.004859179,"threshold_uncertainty_score":0.01391417,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1012755886281797,"score_gpt":0.2261991095555468,"score_spread":0.1249235209273671,"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."}}