{"id":"https://openalex.org/W7130560853","doi":"https://doi.org/10.1109/apccas67402.2025.11376941","title":"CoRN-LN: Compressed Reciprocal Newton Method for Efficient Layer Normalization","display_name":"CoRN-LN: Compressed Reciprocal Newton Method for Efficient Layer Normalization","publication_year":2025,"publication_date":"2025-10-12","ids":{"openalex":"https://openalex.org/W7130560853","doi":"https://doi.org/10.1109/apccas67402.2025.11376941"},"language":null,"primary_location":{"id":"doi:10.1109/apccas67402.2025.11376941","is_oa":false,"landing_page_url":"https://doi.org/10.1109/apccas67402.2025.11376941","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Asia Pacific Conference on Circuits and Systems (APCCAS)","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/A5126433774","display_name":"Dawon Choi","orcid":null},"institutions":[{"id":"https://openalex.org/I4575257","display_name":"Hanyang University","ror":"https://ror.org/046865y68","country_code":"KR","type":"education","lineage":["https://openalex.org/I4575257"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Dawon Choi","raw_affiliation_strings":["Hanyang University,Dept. of Artificial Intelligence Semiconductor Engineering,Seoul,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hanyang University,Dept. of Artificial Intelligence Semiconductor Engineering,Seoul,Republic of Korea","institution_ids":["https://openalex.org/I4575257"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126373573","display_name":"Hana Kim","orcid":null},"institutions":[{"id":"https://openalex.org/I4575257","display_name":"Hanyang University","ror":"https://ror.org/046865y68","country_code":"KR","type":"education","lineage":["https://openalex.org/I4575257"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hana Kim","raw_affiliation_strings":["Hanyang University,Dept. of Electronic Engineering,Seoul,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hanyang University,Dept. of Electronic Engineering,Seoul,Republic of Korea","institution_ids":["https://openalex.org/I4575257"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5002957476","display_name":"J. Kim","orcid":null},"institutions":[{"id":"https://openalex.org/I4575257","display_name":"Hanyang University","ror":"https://ror.org/046865y68","country_code":"KR","type":"education","lineage":["https://openalex.org/I4575257"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Ji-Hoon Kim","raw_affiliation_strings":["Hanyang University,Dept. of Artificial Intelligence Semiconductor Engineering,Seoul,Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hanyang University,Dept. of Artificial Intelligence Semiconductor Engineering,Seoul,Republic of Korea","institution_ids":["https://openalex.org/I4575257"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.57496924,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"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/T10363","display_name":"Low-power high-performance VLSI design","score":0.1826999932527542,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10363","display_name":"Low-power high-performance VLSI design","score":0.1826999932527542,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.14319999516010284,"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"}},{"id":"https://openalex.org/T10904","display_name":"Embedded Systems Design Techniques","score":0.1200999990105629,"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/normalization","display_name":"Normalization (sociology)","score":0.788100004196167},{"id":"https://openalex.org/keywords/reciprocal","display_name":"Reciprocal","score":0.6473000049591064},{"id":"https://openalex.org/keywords/mean-squared-error","display_name":"Mean squared error","score":0.4341999888420105},{"id":"https://openalex.org/keywords/verilog","display_name":"Verilog","score":0.43130001425743103},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.3797999918460846},{"id":"https://openalex.org/keywords/square-root","display_name":"Square root","score":0.3465999960899353},{"id":"https://openalex.org/keywords/offset","display_name":"Offset (computer science)","score":0.34470000863075256},{"id":"https://openalex.org/keywords/exponential-function","display_name":"Exponential function","score":0.3407999873161316}],"concepts":[{"id":"https://openalex.org/C136886441","wikidata":"https://www.wikidata.org/wiki/Q926129","display_name":"Normalization (sociology)","level":2,"score":0.788100004196167},{"id":"https://openalex.org/C2777742833","wikidata":"https://www.wikidata.org/wiki/Q1964083","display_name":"Reciprocal","level":2,"score":0.6473000049591064},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.6074000000953674},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5415999889373779},{"id":"https://openalex.org/C139945424","wikidata":"https://www.wikidata.org/wiki/Q1940696","display_name":"Mean squared error","level":2,"score":0.4341999888420105},{"id":"https://openalex.org/C2779030575","wikidata":"https://www.wikidata.org/wiki/Q827773","display_name":"Verilog","level":3,"score":0.43130001425743103},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.3797999918460846},{"id":"https://openalex.org/C11577676","wikidata":"https://www.wikidata.org/wiki/Q134237","display_name":"Square root","level":2,"score":0.3465999960899353},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.34470000863075256},{"id":"https://openalex.org/C151376022","wikidata":"https://www.wikidata.org/wiki/Q168698","display_name":"Exponential function","level":2,"score":0.3407999873161316},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.33320000767707825},{"id":"https://openalex.org/C85189116","wikidata":"https://www.wikidata.org/wiki/Q374195","display_name":"Newton's method","level":3,"score":0.329800009727478},{"id":"https://openalex.org/C122383733","wikidata":"https://www.wikidata.org/wiki/Q865920","display_name":"Approximation error","level":2,"score":0.32519999146461487},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.2802000045776367},{"id":"https://openalex.org/C160824197","wikidata":"https://www.wikidata.org/wiki/Q2071054","display_name":"Linear approximation","level":3,"score":0.2718000113964081},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.26600000262260437},{"id":"https://openalex.org/C46362747","wikidata":"https://www.wikidata.org/wiki/Q173431","display_name":"CMOS","level":2,"score":0.2578999996185303},{"id":"https://openalex.org/C42747912","wikidata":"https://www.wikidata.org/wiki/Q1048447","display_name":"Multiplicative function","level":2,"score":0.2556000053882599},{"id":"https://openalex.org/C11210021","wikidata":"https://www.wikidata.org/wiki/Q1520713","display_name":"Linearization","level":3,"score":0.25459998846054077}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/apccas67402.2025.11376941","is_oa":false,"landing_page_url":"https://doi.org/10.1109/apccas67402.2025.11376941","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Asia Pacific Conference on Circuits and Systems (APCCAS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321681","display_name":"Ministry of Trade, Industry and Energy","ror":"https://ror.org/008nkqk13"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W2898915279","https://openalex.org/W3163822089","https://openalex.org/W4293025109","https://openalex.org/W4385488484","https://openalex.org/W4389166736","https://openalex.org/W4403919776","https://openalex.org/W4406137499"],"related_works":[],"abstract_inverted_index":{"In":[0],"Transformer-based":[1],"models":[2],"such":[3,40],"as":[4,41],"BERT,":[5],"Layer":[6,72],"Normalization":[7,73],"(LayerNorm)":[8],"accounts":[9],"for":[10,70,145],"a":[11,77,92],"significant":[12,50],"portion":[13],"of":[14],"total":[15],"computations,":[16],"and":[17,43,55,84,115,117,129,137,170],"its":[18,37],"efficiency":[19],"directly":[20],"impacts":[21],"model":[22],"performance.":[23],"Among":[24],"LayerNorm":[25],"components,":[26],"the":[27,101,118,124,130,140,151],"reciprocal":[28],"square":[29,44],"root":[30],"operation":[31],"is":[32,76],"particularly":[33],"challenging":[34],"due":[35],"to":[36,97,109,112,122,164],"non-linear":[38],"operations":[39,47],"division":[42],"root.":[45],"These":[46],"result":[48],"in":[49,134],"accuracy":[51,83,114],"degradation,":[52],"increased":[53],"latency":[54],"high":[56],"area":[57,175],"overhead.":[58],"To":[59],"address":[60],"these":[61],"challenges,":[62],"this":[63],"paper":[64],"proposes":[65],"Compressed":[66],"Reciprocal":[67],"Newton's":[68],"method":[69,96],"efficient":[71],"(CoRN-LN)":[74],"which":[75],"novel":[78],"approximation":[79],"architecture":[80,128],"that":[81,106,150],"improves":[82],"hardware":[85],"area.":[86],"CoRN-LN":[87,127,152],"introduces":[88],"three":[89],"key":[90],"contributions:":[91],"16":[93],"bit":[94],"compression":[95],"reduce":[98],"computational":[99],"complexity,":[100],"Leading-One-Detector":[102],"(LOD)-aware":[103],"initial":[104],"guess":[105],"dynamically":[107],"adapts":[108],"input":[110],"value":[111],"improve":[113],"latency,":[116],"reciprocal-form":[119],"Newton":[120],"Approximation":[121],"eliminate":[123],"divider.":[125],"The":[126],"baseline":[131],"were":[132],"implemented":[133],"Verilog":[135],"HDL":[136],"synthesized":[138],"using":[139,179],"Samsung":[141],"28nm":[142],"CMOS":[143],"process":[144],"fair":[146],"comparison.":[147],"We":[148],"demonstrate":[149],"reduces":[153],"Mean":[154],"Square":[155],"Error":[156],"(MSE)":[157],"from":[158],"<tex":[159,165,171],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[160,166,172],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$5.50":[161],"\\times":[162,168],"10^{-2}$</tex>":[163],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$8.85":[167],"10^{-6}$</tex>,":[169],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$7":[173],"\\times$</tex>":[174],"reduction":[176],"over":[177],"baseline,":[178],"only":[180],"2-cycle.":[181]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-02-20T00:00:00"}