{"id":"https://openalex.org/W4385080022","doi":"https://doi.org/10.1109/iscas46773.2023.10181465","title":"Hardware-efficient Softmax Approximation for Self-Attention Networks","display_name":"Hardware-efficient Softmax Approximation for Self-Attention Networks","publication_year":2023,"publication_date":"2023-05-21","ids":{"openalex":"https://openalex.org/W4385080022","doi":"https://doi.org/10.1109/iscas46773.2023.10181465"},"language":"en","primary_location":{"id":"doi:10.1109/iscas46773.2023.10181465","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iscas46773.2023.10181465","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 IEEE International Symposium on Circuits and Systems (ISCAS)","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/A5014626587","display_name":"Nazim Altar Koca","orcid":null},"institutions":[{"id":"https://openalex.org/I115228651","display_name":"Agency for Science, Technology and Research","ror":"https://ror.org/036wvzt09","country_code":"SG","type":"government","lineage":["https://openalex.org/I115228651"]},{"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"]},{"id":"https://openalex.org/I4210090209","display_name":"Institute of Microelectronics","ror":"https://ror.org/009rw8n36","country_code":"SG","type":"facility","lineage":["https://openalex.org/I115228651","https://openalex.org/I4210090209","https://openalex.org/I91275662"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Nazim Altar Koca","raw_affiliation_strings":["Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research),Singapore","Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research), Singapore","School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research),Singapore","institution_ids":["https://openalex.org/I115228651","https://openalex.org/I4210090209"]},{"raw_affiliation_string":"Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research), Singapore","institution_ids":["https://openalex.org/I115228651","https://openalex.org/I4210090209"]},{"raw_affiliation_string":"School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore","institution_ids":["https://openalex.org/I172675005"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101615706","display_name":"Anh Tuan","orcid":"https://orcid.org/0000-0002-8320-6818"},"institutions":[{"id":"https://openalex.org/I115228651","display_name":"Agency for Science, Technology and Research","ror":"https://ror.org/036wvzt09","country_code":"SG","type":"government","lineage":["https://openalex.org/I115228651"]},{"id":"https://openalex.org/I4210090209","display_name":"Institute of Microelectronics","ror":"https://ror.org/009rw8n36","country_code":"SG","type":"facility","lineage":["https://openalex.org/I115228651","https://openalex.org/I4210090209","https://openalex.org/I91275662"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Anh Tuan Do","raw_affiliation_strings":["Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research),Singapore","Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research), Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research),Singapore","institution_ids":["https://openalex.org/I115228651","https://openalex.org/I4210090209"]},{"raw_affiliation_string":"Institute of Microelectronics, A*STAR (Agency for Science, Technology and Research), Singapore","institution_ids":["https://openalex.org/I115228651","https://openalex.org/I4210090209"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029335324","display_name":"Chip-Hong Chang","orcid":"https://orcid.org/0000-0002-8897-6176"},"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":"Chip-Hong Chang","raw_affiliation_strings":["School of Electrical and Electronic Engineering, Nanyang Technological University,Singapore","School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical and Electronic Engineering, Nanyang Technological University,Singapore","institution_ids":["https://openalex.org/I172675005"]},{"raw_affiliation_string":"School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore","institution_ids":["https://openalex.org/I172675005"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":5.4647,"has_fulltext":false,"cited_by_count":26,"citation_normalized_percentile":{"value":0.96822061,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":98,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"5"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9986000061035156,"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/T10028","display_name":"Topic Modeling","score":0.9986000061035156,"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/T11714","display_name":"Multimodal Machine Learning Applications","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"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9865999817848206,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/softmax-function","display_name":"Softmax function","score":0.9012892842292786},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7754932641983032},{"id":"https://openalex.org/keywords/datapath","display_name":"Datapath","score":0.6621404886245728},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.5558764934539795},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5232669115066528},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.496202290058136},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4665662348270416},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.4221177101135254},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.3941677212715149},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.346097856760025},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.2100083827972412},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.13391247391700745},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.12566152215003967}],"concepts":[{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.9012892842292786},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7754932641983032},{"id":"https://openalex.org/C2781198647","wikidata":"https://www.wikidata.org/wiki/Q1633673","display_name":"Datapath","level":2,"score":0.6621404886245728},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.5558764934539795},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5232669115066528},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.496202290058136},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4665662348270416},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.4221177101135254},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3941677212715149},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.346097856760025},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.2100083827972412},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.13391247391700745},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.12566152215003967},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"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/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iscas46773.2023.10181465","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iscas46773.2023.10181465","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 IEEE International Symposium on Circuits and Systems (ISCAS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":35,"referenced_works":["https://openalex.org/W2608093348","https://openalex.org/W2896457183","https://openalex.org/W2903640996","https://openalex.org/W2923014074","https://openalex.org/W2963748441","https://openalex.org/W2965373594","https://openalex.org/W2980282514","https://openalex.org/W3017024317","https://openalex.org/W3035083896","https://openalex.org/W3036061420","https://openalex.org/W3046387293","https://openalex.org/W3090555870","https://openalex.org/W3159727696","https://openalex.org/W3170233084","https://openalex.org/W3189877953","https://openalex.org/W3207366647","https://openalex.org/W3208397797","https://openalex.org/W3211525823","https://openalex.org/W3216726218","https://openalex.org/W4226250047","https://openalex.org/W4280582299","https://openalex.org/W4292779060","https://openalex.org/W4293023328","https://openalex.org/W4293768835","https://openalex.org/W4385245566","https://openalex.org/W6739901393","https://openalex.org/W6755207826","https://openalex.org/W6766673545","https://openalex.org/W6768817161","https://openalex.org/W6778883912","https://openalex.org/W6779491015","https://openalex.org/W6788001715","https://openalex.org/W6802393872","https://openalex.org/W6804156745","https://openalex.org/W6811409025"],"related_works":["https://openalex.org/W3107204728","https://openalex.org/W4287591324","https://openalex.org/W4226420367","https://openalex.org/W2109699519","https://openalex.org/W2980176872","https://openalex.org/W970262775","https://openalex.org/W2962876041","https://openalex.org/W1831618318","https://openalex.org/W3090555870","https://openalex.org/W2268150819"],"abstract_inverted_index":{"Self-attention":[0],"networks":[1],"such":[2],"as":[3,19,86],"Transformer":[4,38],"have":[5],"become":[6],"state-of-the-art":[7],"models":[8,67],"for":[9,65,158],"natural":[10],"language":[11,142],"processing":[12],"(NLP)":[13],"problems.":[14],"Softmax":[15,40,80,114,157],"function,":[16],"which":[17,51,82],"serves":[18],"a":[20,30,37,77,87,148,174],"normalizer":[21],"to":[22,28,55,95,151,172],"produce":[23],"attention":[24],"scores,":[25],"turns":[26],"out":[27],"be":[29,62,84],"severe":[31],"throughput":[32],"and":[33,48,57,126,134],"latency":[34,129],"bottleneck":[35],"of":[36,43,156,164,178],"network.":[39],"datapath":[41],"consists":[42],"data-dependent":[44],"sequential":[45],"nonlinear":[46],"exponentiation":[47],"division":[49],"operations,":[50],"are":[52],"not":[53],"amenable":[54],"pipelining":[56],"parallelism,":[58],"nor":[59],"can":[60,83],"they":[61],"directly":[63],"linearized":[64],"pretrained":[66,92],"without":[68,99],"substantial":[69],"accuracy":[70,138],"drop.":[71],"In":[72],"this":[73],"paper,":[74],"we":[75],"proposed":[76],"hardware":[78],"efficient":[79],"approximation":[81],"used":[85],"direct":[88],"plug-in":[89],"substitution":[90],"into":[91],"transformer":[93],"network":[94],"accelerate":[96],"NLP":[97,159],"tasks":[98],"compromising":[100],"its":[101],"accuracy.":[102],"Experiment":[103],"results":[104],"on":[105,140,169],"FPGA":[106],"implementation":[107],"show":[108],"that":[109],"our":[110],"design":[111],"outperforms":[112],"vanilla":[113],"designed":[115],"using":[116],"Xilinx":[117],"IPs":[118],"with":[119,161],"15x":[120],"less":[121,124,135],"LUTs,":[122],"55x":[123],"registers":[125],"23x":[127],"lower":[128],"at":[130],"similar":[131],"clock":[132],"frequency":[133],"than":[136],"1%":[137],"drop":[139],"main":[141],"benchmark":[143],"tasks.":[144],"We":[145],"also":[146],"propose":[147],"pruning":[149],"method":[150],"reduce":[152],"the":[153],"input":[154],"entropy":[155],"problems":[160],"high":[162],"number":[163],"inputs.":[165],"It":[166],"was":[167],"validated":[168],"CoLA":[170],"task":[171],"achieve":[173],"further":[175],"25%":[176],"reduction":[177],"latency.":[179]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":17},{"year":2024,"cited_by_count":5}],"updated_date":"2026-08-11T07:18:39.950985","created_date":"2025-10-10T00:00:00"}