{"id":"https://openalex.org/W4400811165","doi":"https://doi.org/10.1109/aicas59952.2024.10595882","title":"Reusing Softmax Hardware Unit for GELU Computation in Transformers","display_name":"Reusing Softmax Hardware Unit for GELU Computation in Transformers","publication_year":2024,"publication_date":"2024-04-22","ids":{"openalex":"https://openalex.org/W4400811165","doi":"https://doi.org/10.1109/aicas59952.2024.10595882"},"language":"en","primary_location":{"id":"doi:10.1109/aicas59952.2024.10595882","is_oa":false,"landing_page_url":"https://doi.org/10.1109/aicas59952.2024.10595882","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2024 IEEE 6th International Conference on AI Circuits and Systems (AICAS)","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/A5063756965","display_name":"Christodoulos Peltekis","orcid":"https://orcid.org/0009-0000-8930-9431"},"institutions":[{"id":"https://openalex.org/I147962203","display_name":"Democritus University of Thrace","ror":"https://ror.org/03bfqnx40","country_code":"GR","type":"education","lineage":["https://openalex.org/I147962203"]}],"countries":["GR"],"is_corresponding":true,"raw_author_name":"Christodoulos Peltekis","raw_affiliation_strings":["Democritus University of Thrace,Electrical and Computer Engineering,Xanthi,Greece"],"affiliations":[{"raw_affiliation_string":"Democritus University of Thrace,Electrical and Computer Engineering,Xanthi,Greece","institution_ids":["https://openalex.org/I147962203"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111262390","display_name":"Kosmas Alexandridis","orcid":null},"institutions":[{"id":"https://openalex.org/I147962203","display_name":"Democritus University of Thrace","ror":"https://ror.org/03bfqnx40","country_code":"GR","type":"education","lineage":["https://openalex.org/I147962203"]}],"countries":["GR"],"is_corresponding":false,"raw_author_name":"Kosmas Alexandridis","raw_affiliation_strings":["Democritus University of Thrace,Electrical and Computer Engineering,Xanthi,Greece"],"affiliations":[{"raw_affiliation_string":"Democritus University of Thrace,Electrical and Computer Engineering,Xanthi,Greece","institution_ids":["https://openalex.org/I147962203"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5074704256","display_name":"Giorgos Dimitrakopoulos","orcid":"https://orcid.org/0000-0003-3688-7865"},"institutions":[{"id":"https://openalex.org/I147962203","display_name":"Democritus University of Thrace","ror":"https://ror.org/03bfqnx40","country_code":"GR","type":"education","lineage":["https://openalex.org/I147962203"]}],"countries":["GR"],"is_corresponding":false,"raw_author_name":"Giorgos Dimitrakopoulos","raw_affiliation_strings":["Democritus University of Thrace,Electrical and Computer Engineering,Xanthi,Greece"],"affiliations":[{"raw_affiliation_string":"Democritus University of Thrace,Electrical and Computer Engineering,Xanthi,Greece","institution_ids":["https://openalex.org/I147962203"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5063756965"],"corresponding_institution_ids":["https://openalex.org/I147962203"],"apc_list":null,"apc_paid":null,"fwci":6.2377,"has_fulltext":false,"cited_by_count":12,"citation_normalized_percentile":{"value":0.97599898,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"159","last_page":"163"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9261000156402588,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9261000156402588,"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/reuse","display_name":"Reuse","score":0.7429613471031189},{"id":"https://openalex.org/keywords/softmax-function","display_name":"Softmax function","score":0.659905731678009},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6045640110969543},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5590100288391113},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.4870198369026184},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.45791900157928467},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.34150853753089905},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.20960238575935364},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.17589512467384338},{"id":"https://openalex.org/keywords/waste-management","display_name":"Waste management","score":0.09753066301345825},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.09505832195281982},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.06538912653923035},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.0641716718673706}],"concepts":[{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.7429613471031189},{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.659905731678009},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6045640110969543},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5590100288391113},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4870198369026184},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.45791900157928467},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.34150853753089905},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.20960238575935364},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.17589512467384338},{"id":"https://openalex.org/C548081761","wikidata":"https://www.wikidata.org/wiki/Q180388","display_name":"Waste management","level":1,"score":0.09753066301345825},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.09505832195281982},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.06538912653923035},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0641716718673706},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/aicas59952.2024.10595882","is_oa":false,"landing_page_url":"https://doi.org/10.1109/aicas59952.2024.10595882","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2024 IEEE 6th International Conference on AI Circuits and Systems (AICAS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy","score":0.5699999928474426}],"awards":[],"funders":[{"id":"https://openalex.org/F4320322923","display_name":"Siemens","ror":"https://ror.org/059mq0909"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":31,"referenced_works":["https://openalex.org/W2151577341","https://openalex.org/W2608093348","https://openalex.org/W2801655600","https://openalex.org/W2896457183","https://openalex.org/W2899663614","https://openalex.org/W2923014074","https://openalex.org/W2963846996","https://openalex.org/W2978670439","https://openalex.org/W3036061420","https://openalex.org/W3046387293","https://openalex.org/W3082834449","https://openalex.org/W3211525823","https://openalex.org/W4288089799","https://openalex.org/W4293768835","https://openalex.org/W4312399146","https://openalex.org/W4322718191","https://openalex.org/W4385080022","https://openalex.org/W4385245566","https://openalex.org/W4386765242","https://openalex.org/W4386859272","https://openalex.org/W6605323724","https://openalex.org/W6691459498","https://openalex.org/W6751068846","https://openalex.org/W6755207826","https://openalex.org/W6755977528","https://openalex.org/W6768021236","https://openalex.org/W6769627184","https://openalex.org/W6784333009","https://openalex.org/W6788001715","https://openalex.org/W6793874019","https://openalex.org/W6850625674"],"related_works":["https://openalex.org/W3107204728","https://openalex.org/W4287591324","https://openalex.org/W3108503355","https://openalex.org/W4226420367","https://openalex.org/W2962876041","https://openalex.org/W3090555870","https://openalex.org/W3022820045","https://openalex.org/W2801655600","https://openalex.org/W3005627584","https://openalex.org/W4303493643"],"abstract_inverted_index":{"Transformers":[0],"have":[1],"improved":[2],"drastically":[3],"the":[4,65,76,97,129,138,141],"performance":[5],"of":[6,17,67,93,101,131,140],"natural":[7],"language":[8],"processing":[9],"(NLP)":[10],"and":[11,22,29,50,103,119,135,145,149],"computer":[12],"vision":[13],"applications.":[14],"The":[15],"computation":[16,66],"transformers":[18],"involves":[19],"matrix":[20],"multiplications":[21],"non-linear":[23],"activation":[24],"functions":[25],"such":[26],"as":[27,90],"softmax":[28,71,83,102,122],"GELU":[30,68,89,94,108,115],"(Gaussion":[31],"Error":[32],"Linear":[33],"Unit)":[34],"that":[35,113],"are":[36],"accelerated":[37],"directly":[38],"in":[39,60,105],"hardware.":[40],"Currently,":[41],"function":[42,49],"evaluation":[43],"is":[44],"done":[45],"separately":[46],"for":[47,53,82,87],"each":[48],"rarely":[51],"allows":[52,137],"hardware":[54,78,123,143],"reuse.":[55],"To":[56],"mitigate":[57],"this":[58,61,74],"problem,":[59],"work,":[62],"we":[63],"map":[64],"to":[69],"a":[70,117],"operator.":[72],"In":[73],"way,":[75],"efficient":[77],"units":[79],"designed":[80],"already":[81],"can":[84,95],"be":[85],"reused":[86],"computing":[88,114],"well.":[91],"Computation":[92],"enjoy":[96],"inherent":[98],"vectorized":[99],"nature":[100],"produce":[104],"parallel":[106],"multiple":[107],"outcomes.":[109],"Experimental":[110],"results":[111],"show":[112],"via":[116],"pre-existing":[118],"incrementally":[120],"modified":[121],"unit":[124],"(a)":[125],"does":[126],"not":[127],"reduce":[128],"accuracy":[130],"representative":[132],"NLP":[133],"applications":[134],"(b)":[136],"reduction":[139],"overall":[142],"area":[144],"power":[146],"by":[147],"6.1%":[148],"11.9%,":[150],"respectively,":[151],"on":[152],"average.":[153]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":9},{"year":2024,"cited_by_count":2}],"updated_date":"2025-12-26T23:08:49.675405","created_date":"2025-10-10T00:00:00"}