{"id":"https://openalex.org/W2915106038","doi":"https://doi.org/10.1145/3289602.3293898","title":"Efficient and Effective Sparse LSTM on FPGA with Bank-Balanced Sparsity","display_name":"Efficient and Effective Sparse LSTM on FPGA with Bank-Balanced Sparsity","publication_year":2019,"publication_date":"2019-02-20","ids":{"openalex":"https://openalex.org/W2915106038","doi":"https://doi.org/10.1145/3289602.3293898","mag":"2915106038"},"language":"en","primary_location":{"id":"doi:10.1145/3289602.3293898","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3289602.3293898","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays","raw_type":"proceedings-article"},"type":"conference-paper","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/A5000260708","display_name":"Shi\u2010Jie Cao","orcid":"https://orcid.org/0000-0001-9136-0949"},"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":"Shijie Cao","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100374175","display_name":"Chen Zhang","orcid":"https://orcid.org/0000-0003-2762-2726"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen Zhang","raw_affiliation_strings":["Microsoft Research, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research, Beijing, China","institution_ids":["https://openalex.org/I4210113369"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012287052","display_name":"Zhuliang Yao","orcid":null},"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":"Zhuliang Yao","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":"middle","author":{"id":"https://openalex.org/A5086945155","display_name":"Wencong Xiao","orcid":"https://orcid.org/0000-0002-3043-522X"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wencong Xiao","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100642163","display_name":"Lanshun Nie","orcid":"https://orcid.org/0000-0002-8354-8359"},"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":"Lanshun Nie","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113748483","display_name":"Dechen Zhan","orcid":null},"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":"Dechen Zhan","raw_affiliation_strings":["Harbin Institute of Technology, Harbin, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Harbin Institute of Technology, Harbin, China","institution_ids":["https://openalex.org/I204983213"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102880548","display_name":"Yunxin Liu","orcid":"https://orcid.org/0000-0001-7352-8955"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yunxin Liu","raw_affiliation_strings":["Microsoft Research, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research, Beijing, China","institution_ids":["https://openalex.org/I4210113369"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101931991","display_name":"Ming Wu","orcid":"https://orcid.org/0000-0002-6993-1431"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ming Wu","raw_affiliation_strings":["Microsoft Research, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research, Beijing, China","institution_ids":["https://openalex.org/I4210113369"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101985940","display_name":"Lintao Zhang","orcid":"https://orcid.org/0009-0005-7527-8183"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Lintao Zhang","raw_affiliation_strings":["Microsoft Research, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research, Beijing, China","institution_ids":["https://openalex.org/I4210113369"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":180,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"63","last_page":"72"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.993399977684021,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.993399977684021,"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/T10320","display_name":"Neural Networks and Applications","score":0.9909999966621399,"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/T12326","display_name":"Network Packet Processing and Optimization","score":0.9902999997138977,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8451957702636719},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.7255990505218506},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5860368609428406},{"id":"https://openalex.org/keywords/sparse-matrix","display_name":"Sparse matrix","score":0.5434326529502869},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.5269489884376526},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.33293619751930237},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.3293679356575012}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8451957702636719},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.7255990505218506},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5860368609428406},{"id":"https://openalex.org/C56372850","wikidata":"https://www.wikidata.org/wiki/Q1050404","display_name":"Sparse matrix","level":3,"score":0.5434326529502869},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5269489884376526},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.33293619751930237},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.3293679356575012},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C6557445","wikidata":"https://www.wikidata.org/wiki/Q173113","display_name":"Agronomy","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:10.1145/3289602.3293898","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3289602.3293898","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2019 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Affordable and clean energy","score":0.9100000262260437,"id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":27,"referenced_works":["https://openalex.org/W1591801644","https://openalex.org/W1635512741","https://openalex.org/W1969102784","https://openalex.org/W2009398166","https://openalex.org/W2027069801","https://openalex.org/W2064675550","https://openalex.org/W2094756095","https://openalex.org/W2124007994","https://openalex.org/W2286365479","https://openalex.org/W2293634267","https://openalex.org/W2515287984","https://openalex.org/W2525778437","https://openalex.org/W2542189141","https://openalex.org/W2565851976","https://openalex.org/W2585720638","https://openalex.org/W2606722458","https://openalex.org/W2619096655","https://openalex.org/W2625954420","https://openalex.org/W2657126969","https://openalex.org/W2788838111","https://openalex.org/W2792028943","https://openalex.org/W2883929540","https://openalex.org/W2962820060","https://openalex.org/W2963000224","https://openalex.org/W2963674932","https://openalex.org/W2964299589","https://openalex.org/W4230989867"],"related_works":["https://openalex.org/W2111241003","https://openalex.org/W4200391368","https://openalex.org/W2355315220","https://openalex.org/W2210979487","https://openalex.org/W2074043759","https://openalex.org/W2373535795","https://openalex.org/W2373300491","https://openalex.org/W2082487009","https://openalex.org/W2378744544","https://openalex.org/W4237139544"],"abstract_inverted_index":{"Neural":[0],"networks":[1,209],"based":[2],"on":[3,49,152,195,206,235,243],"Long":[4],"Short-Term":[5],"Memory":[6],"(LSTM)":[7],"are":[8],"widely":[9],"deployed":[10],"in":[11,38,136,172],"latency-sensitive":[12],"language":[13],"and":[14,35,191,238],"speech":[15],"applications.":[16],"To":[17,51],"speed":[18],"up":[19],"LSTM":[20,41,208,222],"inference,":[21],"previous":[22],"research":[23],"proposes":[24],"weight":[25,67,107,153],"pruning":[26,118,145],"techniques":[27],"to":[28,61,122,133,147,174,187,217],"reduce":[29],"computational":[30],"cost.":[31],"Unfortunately,":[32],"irregular":[33,189],"computation":[34,190],"memory":[36,192],"accesses":[37],"unrestricted":[39],"sparse":[40,66,160,207],"limit":[42],"the":[43,63,149,199,227],"realizable":[44],"parallelism,":[45],"especially":[46],"when":[47],"implemented":[48],"FPGA.":[50],"address":[52],"this":[53],"issue,":[54],"some":[55],"researchers":[56],"propose":[57,142],"block-based":[58],"sparsity":[59,85,95],"patterns":[60],"increase":[62],"regularity":[64],"of":[65,185,214,248],"matrices,":[68],"but":[69],"these":[70],"approaches":[71],"suffer":[72],"from":[73],"deteriorated":[74],"prediction":[75],"accuracy.":[76,125,250],"This":[77],"work":[78],"presents":[79],"Bank-Balanced":[80],"Sparsity":[81],"(BBS),":[82],"a":[83,93,128,143,158,211],"novel":[84],"pattern":[86,151],"that":[87,167,182],"can":[88,202],"maintain":[89,123],"model":[90,124,249],"accuracy":[91],"at":[92],"high":[94],"level":[96],"while":[97,115],"still":[98],"enable":[99],"an":[100,179],"efficient":[101],"FPGA":[102,138,180,219],"implementation.":[103],"BBS":[104,135,150,173,186,200,228],"partitions":[105],"each":[106,120],"matrix":[108,161],"row":[109],"into":[110],"banks":[111],"for":[112,221],"parallel":[113],"computing,":[114],"adopts":[116],"fine-grained":[117],"inside":[119],"bank":[121],"We":[126],"develop":[127],"3-step":[129],"software-hardware":[130],"co-optimization":[131],"approach":[132],"apply":[134],"real":[137],"hardware.":[139,175],"First,":[140],"we":[141,156,177],"bank-balanced":[144],"method":[146],"induce":[148],"matrices.":[154],"Then":[155],"introduce":[157],"decoding-free":[159],"format,":[162],"Compressed":[163],"Sparse":[164],"Banks":[165],"(CSB),":[166],"transparently":[168],"exposes":[169],"inter-bank":[170],"parallelism":[171],"Finally,":[176],"design":[178],"accelerator":[181,201,229],"takes":[183],"advantage":[184],"eliminate":[188],"accesses.":[193],"Implemented":[194],"Intel":[196],"Arria-10":[197],"FPGA,":[198],"achieve":[203],"750.9":[204],"GOPs":[205],"with":[210,223,245],"batch":[212],"size":[213],"1.":[215],"Compared":[216],"state-of-the-art":[218],"accelerators":[220],"different":[224],"compression":[225],"techniques,":[226],"achieves":[230],"2.3":[231],"~":[232,240],"3.7x":[233],"improvement":[234],"energy":[236],"efficiency":[237],"7.0":[239],"34.4x":[241],"reduction":[242],"latency":[244],"negligible":[246],"loss":[247]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":18},{"year":2024,"cited_by_count":15},{"year":2023,"cited_by_count":25},{"year":2022,"cited_by_count":25},{"year":2021,"cited_by_count":47},{"year":2020,"cited_by_count":34},{"year":2019,"cited_by_count":14}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}