{"id":"https://openalex.org/W2994144272","doi":"https://doi.org/10.1109/tbdata.2019.2957478","title":"GradientFlow: Optimizing Network Performance for Large-Scale Distributed DNN Training","display_name":"GradientFlow: Optimizing Network Performance for Large-Scale Distributed DNN Training","publication_year":2019,"publication_date":"2019-01-01","ids":{"openalex":"https://openalex.org/W2994144272","doi":"https://doi.org/10.1109/tbdata.2019.2957478","mag":"2994144272"},"language":"en","primary_location":{"id":"doi:10.1109/tbdata.2019.2957478","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tbdata.2019.2957478","pdf_url":null,"source":{"id":"https://openalex.org/S2491400915","display_name":"IEEE Transactions on Big Data","issn_l":"2332-7790","issn":["2332-7790","2372-2096"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320439","host_organization_name":"IEEE Computer Society","host_organization_lineage":["https://openalex.org/P4310320439","https://openalex.org/P4310319808"],"host_organization_lineage_names":["IEEE Computer Society","Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Big Data","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"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/A5100614090","display_name":"Peng Sun","orcid":"https://orcid.org/0000-0002-1227-5533"},"institutions":[{"id":"https://openalex.org/I4210128910","display_name":"Group Sense (China)","ror":"https://ror.org/036wd5777","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210128910"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Peng Sun","raw_affiliation_strings":["SenseTime, Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime, Hong Kong","institution_ids":["https://openalex.org/I4210128910"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041572550","display_name":"Yonggang Wen","orcid":"https://orcid.org/0000-0002-2751-5114"},"institutions":[{"id":"https://openalex.org/I172675005","display_name":"Nanyang Technological University","ror":"https://ror.org/02e7b5302","country_code":"SG","type":"education","lineage":["https://openalex.org/I172675005"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Yonggang Wen","raw_affiliation_strings":["School of Computer Science and Engineering, Nanyang Technological University, Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Nanyang Technological University, Singapore","institution_ids":["https://openalex.org/I172675005"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045920192","display_name":"Ruobing Han","orcid":"https://orcid.org/0000-0002-3090-3951"},"institutions":[{"id":"https://openalex.org/I4210128910","display_name":"Group Sense (China)","ror":"https://ror.org/036wd5777","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210128910"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ruobing Han","raw_affiliation_strings":["SenseTime, Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime, Hong Kong","institution_ids":["https://openalex.org/I4210128910"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052485028","display_name":"Wansen Feng","orcid":null},"institutions":[{"id":"https://openalex.org/I4210128910","display_name":"Group Sense (China)","ror":"https://ror.org/036wd5777","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210128910"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wansen Feng","raw_affiliation_strings":["SenseTime, Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime, Hong Kong","institution_ids":["https://openalex.org/I4210128910"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5049910854","display_name":"Shengen Yan","orcid":"https://orcid.org/0009-0005-3858-7972"},"institutions":[{"id":"https://openalex.org/I4210128910","display_name":"Group Sense (China)","ror":"https://ror.org/036wd5777","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210128910"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shengen Yan","raw_affiliation_strings":["SenseTime, Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime, Hong Kong","institution_ids":["https://openalex.org/I4210128910"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.5806,"has_fulltext":false,"cited_by_count":41,"citation_normalized_percentile":{"value":0.92193647,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"1"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.998199999332428,"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"}},{"id":"https://openalex.org/T12702","display_name":"Brain Tumor Detection and Classification","score":0.9977999925613403,"subfield":{"id":"https://openalex.org/subfields/2808","display_name":"Neurology"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.9094279408454895},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.8807130455970764},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.6734092235565186},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.6027226448059082},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5733683705329895},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.5022716522216797},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.47962698340415955},{"id":"https://openalex.org/keywords/distributed-computing","display_name":"Distributed computing","score":0.4696449637413025},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.44234853982925415},{"id":"https://openalex.org/keywords/deep-neural-networks","display_name":"Deep neural networks","score":0.42954644560813904},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.42847487330436707},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.37963080406188965},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3779703974723816},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.14833009243011475},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.08962172269821167},{"id":"https://openalex.org/keywords/wireless","display_name":"Wireless","score":0.0863909125328064}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.9094279408454895},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.8807130455970764},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.6734092235565186},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.6027226448059082},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5733683705329895},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5022716522216797},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.47962698340415955},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.4696449637413025},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.44234853982925415},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.42954644560813904},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.42847487330436707},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37963080406188965},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3779703974723816},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.14833009243011475},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.08962172269821167},{"id":"https://openalex.org/C555944384","wikidata":"https://www.wikidata.org/wiki/Q249","display_name":"Wireless","level":2,"score":0.0863909125328064},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tbdata.2019.2957478","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tbdata.2019.2957478","pdf_url":null,"source":{"id":"https://openalex.org/S2491400915","display_name":"IEEE Transactions on Big Data","issn_l":"2332-7790","issn":["2332-7790","2372-2096"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320439","host_organization_name":"IEEE Computer Society","host_organization_lineage":["https://openalex.org/P4310320439","https://openalex.org/P4310319808"],"host_organization_lineage_names":["IEEE Computer Society","Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Big Data","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":56,"referenced_works":["https://openalex.org/W1563686443","https://openalex.org/W1969805974","https://openalex.org/W2070940137","https://openalex.org/W2078945459","https://openalex.org/W2083842231","https://openalex.org/W2084392717","https://openalex.org/W2094233035","https://openalex.org/W2117539524","https://openalex.org/W2186615578","https://openalex.org/W2194775991","https://openalex.org/W2339765813","https://openalex.org/W2407022425","https://openalex.org/W2577427573","https://openalex.org/W2579247884","https://openalex.org/W2606891064","https://openalex.org/W2618530766","https://openalex.org/W2622263826","https://openalex.org/W2763421725","https://openalex.org/W2766164908","https://openalex.org/W2769856846","https://openalex.org/W2774000609","https://openalex.org/W2884711234","https://openalex.org/W2900824371","https://openalex.org/W2962747323","https://openalex.org/W2962911728","https://openalex.org/W3101036738","https://openalex.org/W3124352525","https://openalex.org/W4242634435","https://openalex.org/W4249935458","https://openalex.org/W4289401659","https://openalex.org/W4294641903","https://openalex.org/W4301239768","https://openalex.org/W4302296459","https://openalex.org/W6628377381","https://openalex.org/W6633628269","https://openalex.org/W6637151318","https://openalex.org/W6638783484","https://openalex.org/W6679154944","https://openalex.org/W6679393576","https://openalex.org/W6680402377","https://openalex.org/W6683722107","https://openalex.org/W6684859321","https://openalex.org/W6686509673","https://openalex.org/W6713134421","https://openalex.org/W6739622702","https://openalex.org/W6739693220","https://openalex.org/W6745245109","https://openalex.org/W6745410505","https://openalex.org/W6746419290","https://openalex.org/W6746514494","https://openalex.org/W6746839373","https://openalex.org/W6748645090","https://openalex.org/W6753209298","https://openalex.org/W6754818105","https://openalex.org/W6756009870","https://openalex.org/W6756488470"],"related_works":["https://openalex.org/W2058965144","https://openalex.org/W2164382479","https://openalex.org/W2146343568","https://openalex.org/W2076915000","https://openalex.org/W1966837078","https://openalex.org/W2078379271","https://openalex.org/W2043460294","https://openalex.org/W4210712718","https://openalex.org/W1905398909","https://openalex.org/W4285818394"],"abstract_inverted_index":{"It":[0],"is":[1,20],"important":[2,122],"to":[3,97,115],"scale":[4],"out":[5],"deep":[6],"neural":[7],"network":[8,77,99,117],"(DNN)":[9],"training":[10,14,30,126],"for":[11,27,68],"reducing":[12],"model":[13],"time.":[15],"The":[16],"high":[17],"communication":[18,64,104,114],"overhead":[19],"one":[21],"of":[22,76],"the":[23,131],"major":[24],"performance":[25],"bottlenecks":[26],"distributed":[28,69],"DNN":[29,41,70],"across":[31],"multiple":[32,103],"GPUs.":[33],"Our":[34],"investigations":[35],"have":[36],"shown":[37],"that":[38],"popular":[39],"open-source":[40],"systems":[42],"could":[43,139],"only":[44,120],"achieve":[45,140],"2.5":[46],"speedup":[47,144],"ratio":[48],"on":[49,130],"64":[50],"GPUs":[51],"connected":[52],"by":[53,101,119],"56":[54],"Gbps":[55],"network.":[56],"To":[57],"address":[58],"this":[59],"problem,":[60],"we":[61,81,93],"propose":[62,94],"a":[63,74,107],"backend":[65],"named":[66],"GradientFlow":[67],"training,":[71,86],"and":[72,87,110,128,142],"employ":[73],"set":[75],"optimization":[78],"techniques.":[79],"First,":[80],"integrate":[82],"ring-based":[83],"allreduce,":[84],"mixed-precision":[85],"computation/communication":[88],"overlap":[89],"into":[90,106],"GradientFlow.":[91],"Second,":[92],"lazy":[95],"allreduce":[96],"improve":[98],"throughput":[100],"fusing":[102],"operations":[105],"single":[108],"one,":[109],"design":[111],"coarse-grained":[112],"sparse":[113],"reduce":[116],"traffic":[118],"transmitting":[121],"gradient":[123],"chunks.":[124],"When":[125],"AlexNet":[127],"ResNet-50":[129],"ImageNet":[132],"dataset":[133],"using":[134],"512":[135],"GPUs,":[136],"our":[137],"approach":[138],"410.2":[141],"434.1":[143],"ratio,":[145],"respectively.":[146]},"counts_by_year":[{"year":2025,"cited_by_count":7},{"year":2024,"cited_by_count":5},{"year":2023,"cited_by_count":3},{"year":2022,"cited_by_count":11},{"year":2021,"cited_by_count":7},{"year":2020,"cited_by_count":7},{"year":2019,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}