{"id":"https://openalex.org/W2892234294","doi":null,"title":"BML: A High-performance, Low-cost Gradient Synchronization Algorithm for DML Training","display_name":"BML: A High-performance, Low-cost Gradient Synchronization Algorithm for DML Training","publication_year":2018,"publication_date":"2018-01-01","ids":{"openalex":"https://openalex.org/W2892234294","mag":"2892234294"},"language":"en","primary_location":{"id":"mag:2892234294","doi":null,"is_oa":false,"landing_page_url":"https://papers.nips.cc/paper/7678-bml-a-high-performance-low-cost-gradient-synchronization-algorithm-for-dml-training.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4363606243","display_name":"neural information processing systems","issn_l":"1049-5258","issn":["1049-5258"],"is_oa":false,"is_in_doaj":false,"is_core":true,"listed_in":["cwts-core","norway-2"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"Neural Information Processing Systems","raw_type":null},"volume":null,"type":"article","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101517637","display_name":"Songtao Wang","orcid":"https://orcid.org/0000-0002-2073-8235","observed_orcids":["https://orcid.org/0000-0002-2073-8235"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Songtao Wang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-2073-8235","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100380787","display_name":"Dan Li","orcid":"https://orcid.org/0000-0003-1418-7171","observed_orcids":["https://orcid.org/0000-0003-1418-7171"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dan Li","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-1418-7171","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060417049","display_name":"Cheng Hong Yang","orcid":"https://orcid.org/0000-0001-7821-0030","observed_orcids":["https://orcid.org/0000-0001-7821-0030"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Cheng Yang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-7821-0030","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078941654","display_name":"Jinkun Geng","orcid":"https://orcid.org/0000-0002-6574-8349","observed_orcids":["https://orcid.org/0000-0002-6574-8349"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jinkun Geng","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-6574-8349","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Yanshu Wang","orcid":"https://orcid.org/0000-0001-8473-8165","observed_orcids":["https://orcid.org/0000-0001-8473-8165"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yanshu Wang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-8473-8165","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100328245","display_name":"Shuai Wang","orcid":"https://orcid.org/0000-0001-8897-9476","observed_orcids":["https://orcid.org/0000-0001-8897-9476"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shuai Wang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-8897-9476","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034104790","display_name":"Shu\u2010Tao Xia","orcid":"https://orcid.org/0000-0002-8639-982X","observed_orcids":["https://orcid.org/0000-0002-8639-982X"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shu\u2010Tao Xia","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":null,"display_name":"Jian Wu","orcid":"https://orcid.org/0000-0002-3394-1507","observed_orcids":["https://orcid.org/0000-0002-3394-1507"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jian Wu","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-3394-1507","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I99065089"],"apc_list":null,"apc_paid":null,"fwci":5.3882,"has_fulltext":false,"cited_by_count":24,"citation_normalized_percentile":{"value":0.96559362,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":88,"max":99},"biblio":{"volume":"31","issue":null,"first_page":"4238","last_page":"4248"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.2786000072956085,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.2786000072956085,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10101","display_name":"Cloud Computing and Resource Management","score":0.2653999924659729,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.263700008392334,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/gradient-synchronization","display_name":"gradient synchronization","score":0.9459999799728394},{"id":"https://openalex.org/keywords/distributed-learning","display_name":"distributed learning","score":0.9520000219345093},{"id":"https://openalex.org/keywords/fat-tree-topology","display_name":"fat-tree topology","score":0.9240000247955322},{"id":"https://openalex.org/keywords/network-performance","display_name":"network performance","score":0.6200000047683716},{"id":"https://openalex.org/keywords/parameter-server","display_name":"parameter server","score":0.7710000276565552},{"id":"https://openalex.org/keywords/training-time-reduction","display_name":"training time reduction","score":0.515999972820282},{"id":"https://openalex.org/keywords/lenet-5","display_name":"LeNet-5","score":0.7409999966621399},{"id":"https://openalex.org/keywords/vgg19","display_name":"VGG19","score":1.0}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7993346452713013},{"id":"https://openalex.org/C31395832","wikidata":"https://www.wikidata.org/wiki/Q1318674","display_name":"Testbed","level":2,"score":0.6769916415214539},{"id":"https://openalex.org/C2778562939","wikidata":"https://www.wikidata.org/wiki/Q1298791","display_name":"Synchronization (alternating current)","level":3,"score":0.6766304969787598},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.6182547807693481},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5438651442527771},{"id":"https://openalex.org/C113174947","wikidata":"https://www.wikidata.org/wiki/Q2859736","display_name":"Tree (set theory)","level":2,"score":0.446898490190506},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3619147539138794},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.3490247130393982},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.11260142922401428},{"id":"https://openalex.org/C127162648","wikidata":"https://www.wikidata.org/wiki/Q16858953","display_name":"Channel (broadcasting)","level":2,"score":0.10173872113227844},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"mag:2892234294","doi":null,"is_oa":false,"landing_page_url":"https://papers.nips.cc/paper/7678-bml-a-high-performance-low-cost-gradient-synchronization-algorithm-for-dml-training.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4363606243","display_name":"neural information processing systems","issn_l":"1049-5258","issn":["1049-5258"],"is_oa":false,"is_in_doaj":false,"is_core":true,"listed_in":["cwts-core","norway-2"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"Neural Information Processing Systems","raw_type":null}],"best_oa_location":null,"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":17,"referenced_works":["https://openalex.org/W1498436455","https://openalex.org/W1528469369","https://openalex.org/W1686810756","https://openalex.org/W2057332538","https://openalex.org/W2083842231","https://openalex.org/W2112796928","https://openalex.org/W2126210439","https://openalex.org/W2130531694","https://openalex.org/W2138302120","https://openalex.org/W2143065961","https://openalex.org/W2155861865","https://openalex.org/W2168231600","https://openalex.org/W2186615578","https://openalex.org/W2208374148","https://openalex.org/W2475334473","https://openalex.org/W2626721309","https://openalex.org/W2794670651"],"related_works":["https://openalex.org/W3036987943","https://openalex.org/W3089731223","https://openalex.org/W3206236031","https://openalex.org/W4220716191","https://openalex.org/W2955863409","https://openalex.org/W2887718373","https://openalex.org/W4386276152","https://openalex.org/W4388894100","https://openalex.org/W3038333728","https://openalex.org/W4366393886"],"abstract_inverted_index":{"In":[0,17],"distributed":[1],"machine":[2],"learning":[3],"(DML),":[4],"the":[5,12,37,48,63,73,84,116],"network":[6,30,34,71],"performance":[7,31],"between":[8],"machines":[9],"significantly":[10],"impacts":[11],"speed":[13],"of":[14,46,76,83,90,95,100,120],"iterative":[15],"training.":[16],"this":[18],"paper":[19],"we":[20],"propose":[21],"BML,":[22],"a":[23,58,69,106],"new":[24],"gradient":[25,85],"synchronization":[26,86],"algorithm":[27,53,67],"with":[28,88,108],"higher":[29],"and":[32,102],"lower":[33],"cost":[35],"than":[36],"current":[38],"practice.":[39],"BML":[40,52,79,114],"runs":[41],"on":[42,68,105],"BCube":[43],"network,":[44],"instead":[45],"using":[47],"traditional":[49],"Fat-Tree":[50,70],"topology.":[51],"is":[54,97],"designed":[55],"in":[56],"such":[57],"way":[59],"that,":[60,113],"compared":[61],"to":[62,125],"parameter":[64],"server":[65,77],"(PS)":[66],"connecting":[72],"same":[74],"number":[75,94],"machines,":[78],"achieves":[80],"theoretically":[81],"1/k":[82],"time,":[87],"k/5":[89],"switches":[91],"(the":[92],"typical":[93],"k":[96],"2\u223c4).":[98],"Experiments":[99],"LeNet-5":[101],"VGG-19":[103],"benchmarks":[104],"testbed":[107],"9":[109],"dual-GPU":[110],"servers":[111],"show":[112],"reduces":[115],"job":[117],"completion":[118],"time":[119],"DML":[121],"training":[122],"by":[123],"up":[124],"56.4%.":[126]},"counts_by_year":[{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":2},{"year":2022,"cited_by_count":1},{"year":2021,"cited_by_count":4},{"year":2020,"cited_by_count":13},{"year":2019,"cited_by_count":3}],"updated_date":"2026-11-03T10:50:19.383937","created_date":"2018-09-27T00:00:00"}