{"id":"https://openalex.org/W3047357290","doi":"https://doi.org/10.1109/infocom41043.2020.9155269","title":"Communication-Efficient Distributed Deep Learning with Merged Gradient Sparsification on GPUs","display_name":"Communication-Efficient Distributed Deep Learning with Merged Gradient Sparsification on GPUs","publication_year":2020,"publication_date":"2020-07-01","ids":{"openalex":"https://openalex.org/W3047357290","doi":"https://doi.org/10.1109/infocom41043.2020.9155269","mag":"3047357290"},"language":"en","primary_location":{"id":"doi:10.1109/infocom41043.2020.9155269","is_oa":false,"landing_page_url":"https://doi.org/10.1109/infocom41043.2020.9155269","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE INFOCOM 2020 - IEEE Conference on Computer Communications","raw_type":"proceedings-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/A5016836702","display_name":"Shaohuai Shi","orcid":"https://orcid.org/0000-0002-1418-5160"},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":true,"raw_author_name":"Shaohuai Shi","raw_affiliation_strings":["High-Performance Machine Learning Lab, Hong Kong Baptist University"],"affiliations":[{"raw_affiliation_string":"High-Performance Machine Learning Lab, Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018867983","display_name":"Qiang Wang","orcid":"https://orcid.org/0000-0002-2986-967X"},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Qiang Wang","raw_affiliation_strings":["High-Performance Machine Learning Lab, Hong Kong Baptist University"],"affiliations":[{"raw_affiliation_string":"High-Performance Machine Learning Lab, Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100730785","display_name":"Xiaowen Chu","orcid":"https://orcid.org/0000-0001-9745-4372"},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xiaowen Chu","raw_affiliation_strings":["High-Performance Machine Learning Lab, Hong Kong Baptist University"],"affiliations":[{"raw_affiliation_string":"High-Performance Machine Learning Lab, Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060171629","display_name":"Bo Li","orcid":"https://orcid.org/0000-0003-2955-750X"},"institutions":[{"id":"https://openalex.org/I200769079","display_name":"Hong Kong University of Science and Technology","ror":"https://ror.org/00q4vv597","country_code":"HK","type":"education","lineage":["https://openalex.org/I200769079"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Bo Li","raw_affiliation_strings":["Department of Computer Science and Engineering, The Hong Kong University of Science and Technology"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, The Hong Kong University of Science and Technology","institution_ids":["https://openalex.org/I200769079"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052758633","display_name":"Yang Qin","orcid":"https://orcid.org/0000-0002-1981-1423"},"institutions":[{"id":"https://openalex.org/I204983213","display_name":"Harbin Institute of Technology","ror":"https://ror.org/01yqg2h08","country_code":"CN","type":"education","lineage":["https://openalex.org/I204983213"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yang Qin","raw_affiliation_strings":["Department of Computer Science and Technology, Harbin Institute of Technology (Shenzhen)"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology, Harbin Institute of Technology (Shenzhen)","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033074007","display_name":"Ruihao Liu","orcid":"https://orcid.org/0000-0003-1635-1553"},"institutions":[{"id":"https://openalex.org/I4210121678","display_name":"Block Engineering (United States)","ror":"https://ror.org/02ehf4193","country_code":"US","type":"company","lineage":["https://openalex.org/I4210121678"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ruihao Liu","raw_affiliation_strings":["MassGrid.com, Shenzhen District Block Technology Co., Ltd"],"affiliations":[{"raw_affiliation_string":"MassGrid.com, Shenzhen District Block Technology Co., Ltd","institution_ids":["https://openalex.org/I4210121678"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5003655386","display_name":"Xinxiao Zhao","orcid":"https://orcid.org/0000-0001-6568-3904"},"institutions":[{"id":"https://openalex.org/I4210121678","display_name":"Block Engineering (United States)","ror":"https://ror.org/02ehf4193","country_code":"US","type":"company","lineage":["https://openalex.org/I4210121678"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xinxiao Zhao","raw_affiliation_strings":["MassGrid.com, Shenzhen District Block Technology Co., Ltd"],"affiliations":[{"raw_affiliation_string":"MassGrid.com, Shenzhen District Block Technology Co., Ltd","institution_ids":["https://openalex.org/I4210121678"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5016836702"],"corresponding_institution_ids":["https://openalex.org/I141568987"],"apc_list":null,"apc_paid":null,"fwci":6.8931,"has_fulltext":false,"cited_by_count":65,"citation_normalized_percentile":{"value":0.97497502,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"406","last_page":"415"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9988999962806702,"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"}},"topics":[{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9988999962806702,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.9944999814033508,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9939000010490417,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8150674104690552},{"id":"https://openalex.org/keywords/stochastic-gradient-descent","display_name":"Stochastic gradient descent","score":0.7256186008453369},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.701740026473999},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.6984005570411682},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6422668695449829},{"id":"https://openalex.org/keywords/convergence","display_name":"Convergence (economics)","score":0.47589385509490967},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.40824705362319946},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.40574073791503906},{"id":"https://openalex.org/keywords/distributed-computing","display_name":"Distributed computing","score":0.3352894186973572},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.26148727536201477},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.15395256876945496},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.07157739996910095}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8150674104690552},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.7256186008453369},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.701740026473999},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6984005570411682},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6422668695449829},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.47589385509490967},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.40824705362319946},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.40574073791503906},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3352894186973572},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.26148727536201477},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.15395256876945496},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.07157739996910095},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.0},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.0},{"id":"https://openalex.org/C50522688","wikidata":"https://www.wikidata.org/wiki/Q189833","display_name":"Economic growth","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/infocom41043.2020.9155269","is_oa":false,"landing_page_url":"https://doi.org/10.1109/infocom41043.2020.9155269","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE INFOCOM 2020 - IEEE Conference on Computer Communications","raw_type":"proceedings-article"},{"id":"pmh:oai:repository.ust.hk:1783.1-105126","is_oa":false,"landing_page_url":"http://repository.ust.hk/ir/Record/1783.1-105126","pdf_url":null,"source":{"id":"https://openalex.org/S4306401796","display_name":"Rare & Special e-Zone (The Hong Kong University of Science and Technology)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I200769079","host_organization_name":"Hong Kong University of Science and Technology","host_organization_lineage":["https://openalex.org/I200769079"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"Conference paper"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.47999998927116394,"id":"https://metadata.un.org/sdg/9","display_name":"Industry, innovation and infrastructure"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":75,"referenced_works":["https://openalex.org/W1632114991","https://openalex.org/W1686810756","https://openalex.org/W2108598243","https://openalex.org/W2111713407","https://openalex.org/W2131613942","https://openalex.org/W2166221608","https://openalex.org/W2194775991","https://openalex.org/W2402144811","https://openalex.org/W2405578611","https://openalex.org/W2407022425","https://openalex.org/W2498764059","https://openalex.org/W2535374105","https://openalex.org/W2580688187","https://openalex.org/W2606891064","https://openalex.org/W2617766261","https://openalex.org/W2622263826","https://openalex.org/W2745269232","https://openalex.org/W2766140019","https://openalex.org/W2769644379","https://openalex.org/W2774000609","https://openalex.org/W2775776326","https://openalex.org/W2787998955","https://openalex.org/W2798445803","https://openalex.org/W2884711234","https://openalex.org/W2888206291","https://openalex.org/W2891735383","https://openalex.org/W2899771611","https://openalex.org/W2919115771","https://openalex.org/W2919632853","https://openalex.org/W2953384591","https://openalex.org/W2962700998","https://openalex.org/W2962747323","https://openalex.org/W2962758826","https://openalex.org/W2962786385","https://openalex.org/W2963179579","https://openalex.org/W2963540381","https://openalex.org/W2963664311","https://openalex.org/W2963674387","https://openalex.org/W2963786636","https://openalex.org/W2963803379","https://openalex.org/W2964004663","https://openalex.org/W2964140730","https://openalex.org/W2964324519","https://openalex.org/W2964350391","https://openalex.org/W2966527647","https://openalex.org/W2982475424","https://openalex.org/W2985108934","https://openalex.org/W2989768222","https://openalex.org/W2990061623","https://openalex.org/W2990352720","https://openalex.org/W3089807438","https://openalex.org/W3101036738","https://openalex.org/W4297685247","https://openalex.org/W4301239768","https://openalex.org/W6637373629","https://openalex.org/W6713134421","https://openalex.org/W6713835734","https://openalex.org/W6738460352","https://openalex.org/W6739622702","https://openalex.org/W6739693220","https://openalex.org/W6745362205","https://openalex.org/W6745723224","https://openalex.org/W6746200960","https://openalex.org/W6746719585","https://openalex.org/W6746839373","https://openalex.org/W6748019269","https://openalex.org/W6748645090","https://openalex.org/W6752191696","https://openalex.org/W6753209298","https://openalex.org/W6754187583","https://openalex.org/W6754776447","https://openalex.org/W6756040250","https://openalex.org/W6770379276","https://openalex.org/W6770493259","https://openalex.org/W6770895831"],"related_works":["https://openalex.org/W2595172197","https://openalex.org/W2084856301","https://openalex.org/W2127970246","https://openalex.org/W2885125400","https://openalex.org/W1989889224","https://openalex.org/W4382618745","https://openalex.org/W1973775000","https://openalex.org/W2011430815","https://openalex.org/W4321606653","https://openalex.org/W4401278057"],"abstract_inverted_index":{"Distributed":[0],"synchronous":[1],"stochastic":[2],"gradient":[3,52,99,116,136],"descent":[4],"(SGD)":[5],"algorithms":[6],"are":[7],"widely":[8],"used":[9],"in":[10],"large-scale":[11],"deep":[12,146],"learning":[13],"applications,":[14],"while":[15,42,180],"it":[16,85],"is":[17,32],"known":[18],"that":[19,165],"the":[20,24,27,39,47,77,82,88,94,98,106,128,133,154,176],"communication":[21,40,67],"bottleneck":[22],"limits":[23],"scalability":[25],"of":[26,91,145,160],"distributed":[28,143],"system.":[29],"Gradient":[30],"sparsification":[31,53,92,137,190],"a":[33,192],"promising":[34],"technique":[35],"to":[36,127,152,169],"significantly":[37],"reduce":[38,76],"traffic,":[41],"pipelining":[43,59],"can":[44],"further":[45,131],"overlap":[46],"communications":[48,63,109],"with":[49,139,186,196],"computations.":[50],"However,":[51],"introduces":[54],"extra":[55],"computation":[56,89,114],"time,":[57],"and":[58,93,110,115,122,157],"requires":[60],"many":[61],"layer-wise":[62],"which":[64],"introduce":[65],"significant":[66],"startup":[68,78],"overheads.":[69],"Merging":[70],"gradients":[71],"from":[72],"neighbor":[73],"layers":[74],"could":[75],"overheads,":[79],"but":[80],"on":[81,191],"other":[83],"hand":[84],"would":[86],"increase":[87],"time":[90,96,172],"waiting":[95],"for":[97,142],"computation.":[100],"In":[101],"this":[102],"paper,":[103],"we":[104],"formulate":[105],"trade-off":[107],"between":[108],"computations":[111],"(including":[112],"backward":[113],"sparsification)":[117],"as":[118],"an":[119,124],"optimization":[120],"problem,":[121],"derive":[123],"optimal":[125,134],"solution":[126],"problem.":[129],"We":[130,148],"develop":[132],"merged":[135],"algorithm":[138],"SGD":[140,179,188],"(OMGS-SGD)":[141],"training":[144],"learning.":[147],"conduct":[149],"extensive":[150],"experiments":[151],"verify":[153],"convergence":[155,184],"properties":[156],"scaling":[158],"performance":[159,185],"OMGS-SGD.":[161],"Experimental":[162],"results":[163],"show":[164],"OMGS-SGD":[166],"achieves":[167],"up":[168],"31%":[170],"end-to-end":[171],"efficiency":[173],"improvement":[174],"over":[175],"state-of-the-art":[177],"sparsified":[178],"preserving":[181],"nearly":[182],"consistent":[183],"original":[187],"without":[189],"16-GPU":[193],"cluster":[194],"connected":[195],"1Gbps":[197],"Ethernet.":[198]},"counts_by_year":[{"year":2025,"cited_by_count":8},{"year":2024,"cited_by_count":4},{"year":2023,"cited_by_count":18},{"year":2022,"cited_by_count":10},{"year":2021,"cited_by_count":16},{"year":2020,"cited_by_count":8},{"year":2019,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}