{"id":"https://openalex.org/W3136777779","doi":"https://doi.org/10.1109/bigdata50022.2020.9378165","title":"Acceleration of Large Deep Learning Training with Hybrid GPU Memory Management of Swapping and Re-computing","display_name":"Acceleration of Large Deep Learning Training with Hybrid GPU Memory Management of Swapping and Re-computing","publication_year":2020,"publication_date":"2020-12-10","ids":{"openalex":"https://openalex.org/W3136777779","doi":"https://doi.org/10.1109/bigdata50022.2020.9378165","mag":"3136777779"},"language":"en","primary_location":{"id":"doi:10.1109/bigdata50022.2020.9378165","is_oa":false,"landing_page_url":"https://doi.org/10.1109/bigdata50022.2020.9378165","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2020 IEEE International Conference on Big Data (Big Data)","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/A5102286597","display_name":"Haruki Imai","orcid":null},"institutions":[{"id":"https://openalex.org/I4210145865","display_name":"IBM Research - Tokyo","ror":"https://ror.org/04915qk43","country_code":"JP","type":"facility","lineage":["https://openalex.org/I1341412227","https://openalex.org/I4210114115","https://openalex.org/I4210145865"]}],"countries":["JP"],"is_corresponding":true,"raw_author_name":"Haruki Imai","raw_affiliation_strings":["IBM Research \u2013 Tokyo,Tokyo,Japan"],"affiliations":[{"raw_affiliation_string":"IBM Research \u2013 Tokyo,Tokyo,Japan","institution_ids":["https://openalex.org/I4210145865"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089550246","display_name":"Tung D. Le","orcid":null},"institutions":[{"id":"https://openalex.org/I4210145865","display_name":"IBM Research - Tokyo","ror":"https://ror.org/04915qk43","country_code":"JP","type":"facility","lineage":["https://openalex.org/I1341412227","https://openalex.org/I4210114115","https://openalex.org/I4210145865"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Tung D. Le","raw_affiliation_strings":["IBM Research \u2013 Tokyo,Tokyo,Japan"],"affiliations":[{"raw_affiliation_string":"IBM Research \u2013 Tokyo,Tokyo,Japan","institution_ids":["https://openalex.org/I4210145865"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090981513","display_name":"Yasushi Negishi","orcid":null},"institutions":[{"id":"https://openalex.org/I4210145865","display_name":"IBM Research - Tokyo","ror":"https://ror.org/04915qk43","country_code":"JP","type":"facility","lineage":["https://openalex.org/I1341412227","https://openalex.org/I4210114115","https://openalex.org/I4210145865"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Yasushi Negishi","raw_affiliation_strings":["IBM Research \u2013 Tokyo,Tokyo,Japan"],"affiliations":[{"raw_affiliation_string":"IBM Research \u2013 Tokyo,Tokyo,Japan","institution_ids":["https://openalex.org/I4210145865"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5063074861","display_name":"Kiyokuni Kawachiya","orcid":null},"institutions":[{"id":"https://openalex.org/I4210145865","display_name":"IBM Research - Tokyo","ror":"https://ror.org/04915qk43","country_code":"JP","type":"facility","lineage":["https://openalex.org/I1341412227","https://openalex.org/I4210114115","https://openalex.org/I4210145865"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Kiyokuni Kawachiya","raw_affiliation_strings":["IBM Research \u2013 Tokyo,Tokyo,Japan"],"affiliations":[{"raw_affiliation_string":"IBM Research \u2013 Tokyo,Tokyo,Japan","institution_ids":["https://openalex.org/I4210145865"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5102286597"],"corresponding_institution_ids":["https://openalex.org/I4210145865"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.18632789,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"abs 1802 2271","issue":null,"first_page":"1111","last_page":"1116"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998000264167786,"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":0.9998000264167786,"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/T12292","display_name":"Graph Theory and Algorithms","score":0.9937999844551086,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9933000206947327,"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":0.8877967596054077},{"id":"https://openalex.org/keywords/heuristics","display_name":"Heuristics","score":0.738182008266449},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.6221389770507812},{"id":"https://openalex.org/keywords/central-processing-unit","display_name":"Central processing unit","score":0.6158721446990967},{"id":"https://openalex.org/keywords/graphics-processing-unit","display_name":"Graphics processing unit","score":0.5663731694221497},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5464786291122437},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.5255809426307678},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.5160805583000183},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5140898823738098},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.4632806181907654},{"id":"https://openalex.org/keywords/general-purpose-computing-on-graphics-processing-units","display_name":"General-purpose computing on graphics processing units","score":0.4491340219974518},{"id":"https://openalex.org/keywords/graphics","display_name":"Graphics","score":0.4455222487449646},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.43101391196250916},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3404691219329834},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.2762492895126343},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.25225362181663513},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.2164202630519867},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.1617700159549713},{"id":"https://openalex.org/keywords/computer-graphics","display_name":"Computer graphics (images)","score":0.09711417555809021}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8877967596054077},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.738182008266449},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6221389770507812},{"id":"https://openalex.org/C49154492","wikidata":"https://www.wikidata.org/wiki/Q5300","display_name":"Central processing unit","level":2,"score":0.6158721446990967},{"id":"https://openalex.org/C2779851693","wikidata":"https://www.wikidata.org/wiki/Q183484","display_name":"Graphics processing unit","level":2,"score":0.5663731694221497},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5464786291122437},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.5255809426307678},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.5160805583000183},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5140898823738098},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4632806181907654},{"id":"https://openalex.org/C50630238","wikidata":"https://www.wikidata.org/wiki/Q971505","display_name":"General-purpose computing on graphics processing units","level":3,"score":0.4491340219974518},{"id":"https://openalex.org/C21442007","wikidata":"https://www.wikidata.org/wiki/Q1027879","display_name":"Graphics","level":2,"score":0.4455222487449646},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.43101391196250916},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3404691219329834},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2762492895126343},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.25225362181663513},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2164202630519867},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.1617700159549713},{"id":"https://openalex.org/C121684516","wikidata":"https://www.wikidata.org/wiki/Q7600677","display_name":"Computer graphics (images)","level":1,"score":0.09711417555809021},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/bigdata50022.2020.9378165","is_oa":false,"landing_page_url":"https://doi.org/10.1109/bigdata50022.2020.9378165","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2020 IEEE International Conference on Big Data (Big Data)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.6299999952316284,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":18,"referenced_works":["https://openalex.org/W1967134278","https://openalex.org/W2338908902","https://openalex.org/W2489529491","https://openalex.org/W2552911551","https://openalex.org/W2785745756","https://openalex.org/W2913069204","https://openalex.org/W2947664806","https://openalex.org/W2954698599","https://openalex.org/W2964132039","https://openalex.org/W2964174152","https://openalex.org/W2979858475","https://openalex.org/W3008725350","https://openalex.org/W3008884884","https://openalex.org/W3103894541","https://openalex.org/W4301361180","https://openalex.org/W6703652217","https://openalex.org/W6748313825","https://openalex.org/W6763650751"],"related_works":["https://openalex.org/W2983282793","https://openalex.org/W1973046741","https://openalex.org/W1963859303","https://openalex.org/W2364044215","https://openalex.org/W2389600408","https://openalex.org/W240129890","https://openalex.org/W3048701459","https://openalex.org/W2149078538","https://openalex.org/W2080146221","https://openalex.org/W1972148443"],"abstract_inverted_index":{"Deep":[0],"learning":[1],"has":[2],"achieved":[3],"overwhelmingly":[4],"better":[5],"accuracy":[6],"than":[7],"existing":[8,251],"methods":[9],"in":[10,87,137,193,203,227,234,242],"various":[11],"fields.":[12],"To":[13],"further":[14],"improve":[15],"deep":[16],"learning,":[17],"a":[18,146],"deeper":[19],"and":[20,92,104,158,189,240],"larger":[21],"neural":[22],"network":[23],"(NN)":[24],"model":[25],"is":[26,34,51,97,208],"indispensable,":[27],"but":[28,121],"graphics":[29],"processing":[30,71],"unit":[31,72],"(GPU)":[32],"memory":[33,49,60,68,74],"not":[35,79,123],"large":[36,151],"enough":[37],"to":[38,46,69,210,225],"train":[39],"such":[40],"NN":[41,152],"models.":[42],"One":[43],"promising":[44],"method":[45,83,96,117,149,163,220],"reduce":[47,130,211],"GPU":[48,59,67,91],"consumption":[50],"data":[52,65,77,89,102,202],"swapping,":[53],"which":[54,99,174],"eases":[55],"the":[56,76,112,115,138,156,159,165,177,185,191,194,198,204,250],"burden":[57],"on":[58],"by":[61,133,154,214,223],"swapping":[62,201],"out":[63],"intermediate":[64,101],"from":[66],"central":[70],"(CPU)":[73],"while":[75],"are":[78],"necessary.":[80,110],"However,":[81],"this":[82,142],"introduces":[84,118],"communication":[85,132],"overhead":[86],"transferring":[88],"between":[90],"CPU":[93],"memory.":[94],"Another":[95],"recomputation,":[98],"discards":[100],"once":[103],"then":[105],"computes":[106],"them":[107],"again":[108],"when":[109],"Unlike":[111],"data-swapping":[113,139,160,252],"method,":[114],"re-computation":[116,157,192],"additional":[119],"computation":[120],"does":[122],"require":[124],"CPU-GPU":[125,131],"communication.":[126],"Therefore,":[127],"it":[128,135],"may":[129],"introducing":[134],"effectively":[136],"method.":[140,253],"In":[141],"paper,":[143],"we":[144],"developed":[145,171],"faster":[147],"training":[148],"for":[150],"models":[153],"combining":[155],"methods.":[161],"This":[162,206],"edits":[164],"graph":[166,178,186],"of":[167,176,197,200,231,238,246],"TensorFlow":[168],"automatically.":[169],"We":[170],"heuristics":[172,183,207],"about":[173],"part":[175],"should":[179],"be":[180],"recomputed.":[181],"The":[182],"divides":[184],"into":[187],"sub-graphs":[188],"applies":[190],"decreasing":[195],"order":[196],"amount":[199],"sub-graphs.":[205,217],"intended":[209],"maximal":[212],"communications":[213],"re-computing":[215],"minimal":[216],"Our":[218],"hybrid":[219],"improves":[221],"performance":[222],"up":[224],"15.5%":[226],"image":[228,235,243],"size":[229,236,244],"8000\u00d78000":[230],"ResNet50,":[232],"15.3%":[233],"7500\u00d77500":[237],"ResNet152,":[239],"12.4%":[241],"1000\u00d71000":[245],"DeepLabV3+":[247],"compared":[248],"with":[249]},"counts_by_year":[],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}