{"id":"https://openalex.org/W3035430964","doi":"https://doi.org/10.1109/tc.2020.3000118","title":"FPDeep: Scalable Acceleration of CNN Training on Deeply-Pipelined FPGA Clusters","display_name":"FPDeep: Scalable Acceleration of CNN Training on Deeply-Pipelined FPGA Clusters","publication_year":2020,"publication_date":"2020-01-01","ids":{"openalex":"https://openalex.org/W3035430964","doi":"https://doi.org/10.1109/tc.2020.3000118","mag":"3035430964"},"language":"en","primary_location":{"id":"doi:10.1109/tc.2020.3000118","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2020.3000118","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["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 Computers","raw_type":"journal-article"},"type":"article","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/1901.01007","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Tianqi Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I126520041","display_name":"University of Science and Technology of China","ror":"https://ror.org/04c4dkn09","country_code":"CN","type":"education","lineage":["https://openalex.org/I126520041","https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Tianqi Wang","raw_affiliation_strings":["Department of Physics, University of Science and Technology of China, Hefei, China"],"affiliations":[{"raw_affiliation_string":"Department of Physics, University of Science and Technology of China, Hefei, China","institution_ids":["https://openalex.org/I126520041"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Tong Geng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tong Geng","raw_affiliation_strings":["Department of Electrical and Computer Engineering, Boston University, Boston, USA"],"affiliations":[{"raw_affiliation_string":"Department of Electrical and Computer Engineering, Boston University, Boston, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ang Li","orcid":null},"institutions":[{"id":"https://openalex.org/I142606810","display_name":"Pacific Northwest National Laboratory","ror":"https://ror.org/05h992307","country_code":"US","type":"facility","lineage":["https://openalex.org/I1325736334","https://openalex.org/I1330989302","https://openalex.org/I142606810","https://openalex.org/I39565521"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ang Li","raw_affiliation_strings":["PCSD, Pacific Northwest National Laboratory, Richland, USA"],"affiliations":[{"raw_affiliation_string":"PCSD, Pacific Northwest National Laboratory, Richland, USA","institution_ids":["https://openalex.org/I142606810"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Xi Jin","orcid":null},"institutions":[{"id":"https://openalex.org/I126520041","display_name":"University of Science and Technology of China","ror":"https://ror.org/04c4dkn09","country_code":"CN","type":"education","lineage":["https://openalex.org/I126520041","https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xi Jin","raw_affiliation_strings":["Department of Physics, University of Science and Technology of China, Hefei, China"],"affiliations":[{"raw_affiliation_string":"Department of Physics, University of Science and Technology of China, Hefei, China","institution_ids":["https://openalex.org/I126520041"]}]},{"author_position":"last","author":{"id":null,"display_name":"Martin Herbordt","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Martin Herbordt","raw_affiliation_strings":["Department of Electrical and Computer Engineering, Boston University, Boston, USA"],"affiliations":[{"raw_affiliation_string":"Department of Electrical and Computer Engineering, Boston University, Boston, USA","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I126520041"],"apc_list":null,"apc_paid":null,"fwci":1.2758,"has_fulltext":false,"cited_by_count":18,"citation_normalized_percentile":{"value":0.82528338,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":93,"max":98},"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":0.7745000123977661,"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":0.7745000123977661,"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/T14347","display_name":"Big Data and Digital Economy","score":0.05730000138282776,"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/T10273","display_name":"IoT and Edge/Fog Computing","score":0.020999999716877937,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/scalability","display_name":"Scalability","score":0.7694000005722046},{"id":"https://openalex.org/keywords/node","display_name":"Node (physics)","score":0.5539000034332275},{"id":"https://openalex.org/keywords/workload","display_name":"Workload","score":0.5468000173568726},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.5304999947547913},{"id":"https://openalex.org/keywords/convolution","display_name":"Convolution (computer science)","score":0.49309998750686646},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.49239999055862427},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4837000072002411},{"id":"https://openalex.org/keywords/convolutional-neural-network","display_name":"Convolutional neural network","score":0.46549999713897705},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.4490000009536743},{"id":"https://openalex.org/keywords/dram","display_name":"Dram","score":0.34549999237060547}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8773999810218811},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.7694000005722046},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6567999720573425},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.5756999850273132},{"id":"https://openalex.org/C62611344","wikidata":"https://www.wikidata.org/wiki/Q1062658","display_name":"Node (physics)","level":2,"score":0.5539000034332275},{"id":"https://openalex.org/C2778476105","wikidata":"https://www.wikidata.org/wiki/Q628539","display_name":"Workload","level":2,"score":0.5468000173568726},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5304999947547913},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.49309998750686646},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.49239999055862427},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4837000072002411},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.46549999713897705},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.4490000009536743},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.34549999237060547},{"id":"https://openalex.org/C83283714","wikidata":"https://www.wikidata.org/wiki/Q121117","display_name":"Supercomputer","level":2,"score":0.3336000144481659},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3246999979019165},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3228999972343445},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.320499986410141},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.3089999854564667},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.30730000138282776},{"id":"https://openalex.org/C164866538","wikidata":"https://www.wikidata.org/wiki/Q367351","display_name":"Cluster (spacecraft)","level":2,"score":0.305400013923645},{"id":"https://openalex.org/C24885549","wikidata":"https://www.wikidata.org/wiki/Q339678","display_name":"Distributed data store","level":2,"score":0.296099990606308},{"id":"https://openalex.org/C138959212","wikidata":"https://www.wikidata.org/wiki/Q1806783","display_name":"Load balancing (electrical power)","level":3,"score":0.2955000102519989},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.2921000123023987},{"id":"https://openalex.org/C73555534","wikidata":"https://www.wikidata.org/wiki/Q622825","display_name":"Cluster analysis","level":2,"score":0.28839999437332153},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.28290000557899475},{"id":"https://openalex.org/C190475519","wikidata":"https://www.wikidata.org/wiki/Q544384","display_name":"Massively parallel","level":2,"score":0.28220000863075256},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.2809999883174896},{"id":"https://openalex.org/C19012869","wikidata":"https://www.wikidata.org/wiki/Q578372","display_name":"Response time","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.27950000762939453},{"id":"https://openalex.org/C106515295","wikidata":"https://www.wikidata.org/wiki/Q26806595","display_name":"Parallel processing","level":2,"score":0.27889999747276306},{"id":"https://openalex.org/C29140674","wikidata":"https://www.wikidata.org/wiki/Q206637","display_name":"Computer cluster","level":2,"score":0.2777000069618225},{"id":"https://openalex.org/C141917322","wikidata":"https://www.wikidata.org/wiki/Q1025017","display_name":"Cache coherence","level":5,"score":0.27720001339912415},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.27469998598098755},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.27129998803138733},{"id":"https://openalex.org/C34165917","wikidata":"https://www.wikidata.org/wiki/Q188267","display_name":"Programming paradigm","level":2,"score":0.26969999074935913},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.2687999904155731},{"id":"https://openalex.org/C130120984","wikidata":"https://www.wikidata.org/wiki/Q2835898","display_name":"Distributed algorithm","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C2781215313","wikidata":"https://www.wikidata.org/wiki/Q3493345","display_name":"SPARK (programming language)","level":2,"score":0.2678999900817871},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.26030001044273376}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/tc.2020.3000118","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2020.3000118","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["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 Computers","raw_type":"journal-article"},{"id":"pmh:oai:arXiv.org:1901.01007","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1901.01007","pdf_url":"https://arxiv.org/pdf/1901.01007","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:1901.01007","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1901.01007","pdf_url":"https://arxiv.org/pdf/1901.01007","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":43,"referenced_works":["https://openalex.org/W2036779190","https://openalex.org/W2060393849","https://openalex.org/W2155893237","https://openalex.org/W2475840367","https://openalex.org/W2559305746","https://openalex.org/W2597692291","https://openalex.org/W2625954420","https://openalex.org/W2727238169","https://openalex.org/W2729080111","https://openalex.org/W2764337919","https://openalex.org/W2768797272","https://openalex.org/W2769918628","https://openalex.org/W2787766461","https://openalex.org/W2788193959","https://openalex.org/W2798956872","https://openalex.org/W2883929540","https://openalex.org/W2890068895","https://openalex.org/W2903298089","https://openalex.org/W2904445802","https://openalex.org/W2949744956","https://openalex.org/W2969388332","https://openalex.org/W2971542763","https://openalex.org/W2983971024","https://openalex.org/W2984035546","https://openalex.org/W4239385313","https://openalex.org/W4242495425","https://openalex.org/W6684191040","https://openalex.org/W6686509673","https://openalex.org/W6693859313","https://openalex.org/W6694517276","https://openalex.org/W6703420464","https://openalex.org/W6726983090","https://openalex.org/W6727690538","https://openalex.org/W6729480787","https://openalex.org/W6730248441","https://openalex.org/W6734027644","https://openalex.org/W6738144653","https://openalex.org/W6738250615","https://openalex.org/W6739622702","https://openalex.org/W6747917562","https://openalex.org/W6750269818","https://openalex.org/W6753115359","https://openalex.org/W6756718674"],"related_works":[],"abstract_inverted_index":{"Deep":[0],"convolutional":[1],"Neural":[2],"Networks":[3],"(CNNs)":[4],"have":[5],"revolutionized":[6],"numerous":[7,104],"applications,":[8],"but":[9],"the":[10,41,50,57,71,107,137,140,161,178,198,207,215,225,235,255,259],"demand":[11,194],"for":[12,188,206,223],"ever":[13],"more":[14],"performance":[15,113,279],"remains":[16],"unabated.":[17],"Scaling":[18],"CNN":[19],"computations":[20],"to":[21,48,60,93,98,116,147,170,183,197,248,283,292],"larger":[22],"clusters":[23,97],"is":[24,46,139,145,164,195,204,221,271,288],"generally":[25],"done":[26],"by":[27,274],"distributing":[28],"tasks":[29],"in":[30,70,151,156],"batch":[31,117],"mode":[32],"using":[33],"methods":[34],"such":[35],"as":[36],"distributed":[37,51,59,95],"synchronous":[38],"SGD.":[39],"Among":[40],"issues":[42],"with":[43,54,234,254,263,290],"this":[44,66,76],"approach":[45,102],"that,":[47],"make":[49],"cluster":[52],"work":[53,121],"high":[55,171],"utilization,":[56],"workload":[58,130],"each":[61],"node":[62],"must":[63],"be":[64,184],"large;":[65],"implies":[67],"nontrivial":[68],"growth":[69],"SGD":[72],"mini-batch":[73],"size.":[74],"In":[75],"article":[77],"we":[78,212],"propose":[79],"a":[80,86,165,191,218,249],"framework,":[81],"called":[82],"FPDeep,":[83],"which":[84,270],"uses":[85],"hybrid":[87],"of":[88,136,252],"model":[89],"and":[90,122,131,173,175,238],"layer":[91],"parallelism":[92,172],"configure":[94],"reconfigurable":[96],"train":[99],"CNNs.":[100],"This":[101,168],"has":[103,245],"benefits.":[105],"First,":[106],"design":[108],"does":[109],"not":[110],"suffer":[111],"from":[112],"loss":[114],"due":[115],"size":[118],"growth.":[119],"Second,":[120],"storage":[123,193],"are":[124],"balanced":[125],"among":[126],"nodes":[127],"through":[128],"novel":[129],"weight":[132],"partitioning":[133],"schemes.":[134],"Part":[135],"mechanism":[138],"surprising":[141],"finding":[142],"that":[143,180,214,243],"it":[144],"preferable":[146],"store":[148],"excess":[149],"weights":[150],"neighboring":[152],"devices":[153],"rather":[154],"than":[155,302],"local":[157],"off-chip":[158],"memory.":[159],"Third,":[160],"entire":[162],"system":[163],"fine-grained":[166],"pipeline.":[167],"leads":[169],"utilization":[174],"also":[176],"minimizes":[177],"time":[179],"features":[181],"need":[182],"cached":[185],"while":[186],"waiting":[187],"back-propagation.":[189],"As":[190],"result,":[192],"reduced":[196],"point":[199],"where":[200],"only":[201],"on-chip":[202],"memory":[203],"used":[205],"convolution":[208],"layers.":[209],"And":[210],"fourth,":[211],"find":[213],"simplest":[216],"topology,":[217],"1D":[219],"array,":[220],"preferred":[222],"interconnecting":[224],"FPGAs":[226],"thus":[227],"enabling":[228],"widespread":[229],"applicability.":[230],"We":[231],"evaluate":[232],"FPDeep":[233,244,278,294],"Alexnet,":[236],"VGG-16,":[237],"VGG-19":[239],"benchmarks.":[240],"Results":[241],"show":[242],"good":[246],"scalability":[247],"large":[250],"number":[251],"FPGAs,":[253,277],"limiting":[256],"factor":[257],"being":[258],"FPGA-to-FPGA":[260],"bandwidth.":[261],"But":[262],"250":[264],"Gb/s":[265],"bidirectional":[266],"bandwidth":[267],"per":[268],"FPGA,":[269],"easily":[272],"supported":[273],"current":[275],"generation":[276],"shows":[280],"linearity":[281],"up":[282],"100":[284],"FPGAs.":[285],"Energy":[286],"efficiency":[287,301],"evaluated":[289],"respect":[291],"GOPs/J.":[293],"provides,":[295],"on":[296],"average,":[297],"6.4\u00d7":[298],"higher":[299],"energy":[300],"comparable":[303],"GPU":[304],"servers.":[305]},"counts_by_year":[{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":2},{"year":2023,"cited_by_count":6},{"year":2022,"cited_by_count":3},{"year":2021,"cited_by_count":2},{"year":2020,"cited_by_count":2}],"updated_date":"2026-03-20T23:20:44.827607","created_date":"2020-06-19T00:00:00"}