{"id":"W4221026608","doi":"10.1007/s11227-022-04399-2","title":"Heterogeneous gradient computing optimization for scalable deep neural networks","year":2022,"lang":"en","type":"article","venue":"The Journal of Supercomputing","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Agencia Estatal de Investigación; Universidad de Extremadura; Horizon 2020; European Commission; European Regional Development Fund; Banco Bilbao Vizcaya Argentaria; Consejería de Educación y Empleo, Junta de Extremadura; Fundación BBVA; Ministerio de Ciencia e Innovación; Canadian Institute for Advanced Research; Junta de Extremadura; Ministerio de Ciencia, Innovación y Universidades","keywords":"Computer science; Scalability; Replica; Distributed computing; Workload; Artificial neural network; Computation; Partition (number theory); Symmetric multiprocessor system; Data parallelism; Supercomputer; Parallel computing; Process (computing); Artificial intelligence; Parallelism (grammar); Algorithm; Database","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":[0.0009902224,0.001085268,0.001233061,0.0004666336,0.0003902064,0.0007021172,0.001167769,0.000731432,0.001573407],"category_scores_gemma":[0.002142266,0.0004235881,0.0005331829,0.0005981928,0.0006477424,0.0008534917,0.001016149,0.000922174,0.0002514832],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001269684,"about_ca_system_score_gemma":0.001170473,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01001287,"about_ca_topic_score_gemma":0.007431644,"domain_scores_codex":[0.9996558,0.0001102351,0.00001573208,0.00007077623,0.00008623753,0.00006108907],"domain_scores_gemma":[0.9994434,0.0002814437,0.00004362102,0.00004908678,0.0001353881,0.0000471479],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00008959421,0.00004330688,0.0002666893,0.00003366506,0.00002694533,0.00003249238,0.00001274283,0.9794593,0.001197711,0.001907051,0.0006622901,0.01626821],"study_design_scores_gemma":[0.000003375191,0.000005251532,0.00002089674,7.824443e-7,9.656574e-7,9.006258e-7,0.000001222038,0.9993894,0.0001510156,0.0003848367,0.00004070914,6.300028e-7],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1267118,0.001054834,0.865642,0.0005515196,0.0001456088,0.0001132497,0.0001655898,0.001206987,0.004408491],"genre_scores_gemma":[0.837474,0.0002180272,0.1592255,0.0001388422,0.00006071135,0.0001522253,0.0002712078,0.0002457982,0.002213655],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01001287,"threshold_uncertainty_score":0.01990914,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01791124502814537,"score_gpt":0.2477296903838652,"score_spread":0.2298184453557198,"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."}}