{"id":"https://openalex.org/W3209628786","doi":"https://doi.org/10.48550/arxiv.2110.15766","title":"NxMTransformer: Semi-Structured Sparsification for Natural Language Understanding via ADMM","display_name":"NxMTransformer: Semi-Structured Sparsification for Natural Language Understanding via ADMM","publication_year":2021,"publication_date":"2021-10-28","ids":{"openalex":"https://openalex.org/W3209628786","doi":"https://doi.org/10.48550/arxiv.2110.15766","mag":"3209628786"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2110.15766","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2110.15766","pdf_url":"https://arxiv.org/pdf/2110.15766","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2110.15766","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5106543651","display_name":"Connor Holmes","orcid":"https://orcid.org/0000-0002-8314-3677","observed_orcids":["https://orcid.org/0000-0002-8314-3677"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]},{"id":"https://openalex.org/I167576493","display_name":"Colorado School of Mines","ror":"https://ror.org/04raf6v53","country_code":"US","type":"education","lineage":["https://openalex.org/I167576493"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Holmes, Connor","raw_affiliation_strings":["Colorado School of Mines Golden , CO 80401","Yuxiong He Microsoft Bellevue WA , 98004"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Colorado School of Mines Golden , CO 80401","institution_ids":["https://openalex.org/I167576493"]},{"raw_affiliation_string":"Yuxiong He Microsoft Bellevue WA , 98004","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077768924","display_name":"Minjia Zhang","orcid":"https://orcid.org/0000-0002-8165-166X","observed_orcids":["https://orcid.org/0000-0002-8165-166X"]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhang, Minjia","raw_affiliation_strings":["Microsoft Bellevue , WA 98004"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Bellevue , WA 98004","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040302174","display_name":"Yuxiong He","orcid":"https://orcid.org/0000-0003-0478-8854","observed_orcids":["https://orcid.org/0000-0003-0478-8854"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Yuxiong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5059368105","display_name":"Bo Wu","orcid":"https://orcid.org/0000-0001-6658-6452","observed_orcids":["https://orcid.org/0000-0001-6658-6452"]},"institutions":[{"id":"https://openalex.org/I167576493","display_name":"Colorado School of Mines","ror":"https://ror.org/04raf6v53","country_code":"US","type":"education","lineage":["https://openalex.org/I167576493"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wu, Bo","raw_affiliation_strings":["Colorado School of Mines Golden , CO 80401"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Colorado School of Mines Golden , CO 80401","institution_ids":["https://openalex.org/I167576493"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":7,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9452999830245972,"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":0.9452999830245972,"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/T10028","display_name":"Topic Modeling","score":0.04500000178813934,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.002300000051036477,"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/transformer","display_name":"transformer models","score":0.5860000252723694},{"id":"https://openalex.org/keywords/natural-language-understanding","display_name":"natural language understanding","score":0.7110000252723694},{"id":"https://openalex.org/keywords/alternating-direction-method-of-multipliers","display_name":"alternating direction method of multipliers","score":0.7870000004768372},{"id":"https://openalex.org/keywords/pre-trained-language-models","display_name":"pre-trained language models","score":0.8519999980926514},{"id":"https://openalex.org/keywords/constrained-optimization","display_name":"constrained optimization","score":0.824999988079071},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"hardware acceleration","score":0.3790000081062317},{"id":"https://openalex.org/keywords/knowledge-distillation","display_name":"knowledge distillation","score":0.9440000057220459}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8263034820556641},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6486701965332031},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5601434111595154},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.49410679936408997},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4923313856124878},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4739263653755188},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.45845431089401245},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.44153591990470886},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4154786765575409},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3496754467487335},{"id":"https://openalex.org/C6557445","wikidata":"https://www.wikidata.org/wiki/Q173113","display_name":"Agronomy","level":1,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"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/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/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"pmh:oai:arXiv.org:2110.15766","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2110.15766","pdf_url":"https://arxiv.org/pdf/2110.15766","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},{"id":"mag:3213391680","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2110.15766","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"arXiv (Cornell University)","raw_type":null},{"id":"doi:10.48550/arxiv.2110.15766","doi":"https://doi.org/10.48550/arxiv.2110.15766","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2110.15766","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2110.15766","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2110.15766","pdf_url":"https://arxiv.org/pdf/2110.15766","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2271036557","display_name":"SPX: Collaborative Research: Pinpointing and Resolving Scalability Culprits Hidden in Different Components of the Whole System Stack","funder_award_id":"1823005","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G6671297155","display_name":null,"funder_award_id":"CAREER","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G7881724769","display_name":"CAREER: Compiler and Runtime Support for Multi-Tasking on Commodity GPUs","funder_award_id":"1750760","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W3209628786.pdf","grobid_xml":"https://content.openalex.org/works/W3209628786.grobid-xml"},"referenced_works_count":29,"referenced_works":["https://openalex.org/W1821462560","https://openalex.org/W2513419314","https://openalex.org/W2606722458","https://openalex.org/W2621285457","https://openalex.org/W2804032941","https://openalex.org/W2896457183","https://openalex.org/W2897824775","https://openalex.org/W2962818002","https://openalex.org/W2963310665","https://openalex.org/W2963813662","https://openalex.org/W2964121744","https://openalex.org/W2969515962","https://openalex.org/W2970565456","https://openalex.org/W2970971581","https://openalex.org/W2972413484","https://openalex.org/W2976833415","https://openalex.org/W2978017171","https://openalex.org/W2997710335","https://openalex.org/W2998183051","https://openalex.org/W2998342322","https://openalex.org/W3034457371","https://openalex.org/W3035296331","https://openalex.org/W3082274269","https://openalex.org/W3091818438","https://openalex.org/W3105966348","https://openalex.org/W3115511229","https://openalex.org/W3127067080","https://openalex.org/W3129831491","https://openalex.org/W3173374050"],"related_works":["https://openalex.org/W4320085949","https://openalex.org/W4404317337","https://openalex.org/W4291653336","https://openalex.org/W4389982947","https://openalex.org/W4312772368","https://openalex.org/W4382765566","https://openalex.org/W4402670878","https://openalex.org/W3211658005","https://openalex.org/W4226075153","https://openalex.org/W3172288629"],"abstract_inverted_index":{"Natural":[0],"Language":[1],"Processing":[2],"(NLP)":[3],"has":[4],"recently":[5],"achieved":[6],"success":[7],"by":[8],"using":[9],"huge":[10],"pre-trained":[11,87],"Transformer":[12,192],"networks.":[13],"However,":[14,79],"these":[15],"models":[16,132],"often":[17,94],"contain":[18],"hundreds":[19],"of":[20,25,50,57,73,86,160,216],"millions":[21],"or":[22],"even":[23],"billions":[24],"parameters,":[26],"bringing":[27],"challenges":[28],"to":[29,33,46,67,81,96,124,137,145,163,201,212,225],"online":[30],"deployment":[31],"due":[32,80],"latency":[34],"constraints.":[35],"Recently,":[36],"hardware":[37,42,172],"manufacturers":[38],"have":[39,103],"introduced":[40],"dedicated":[41],"for":[43,133,254],"NxM":[44,60,126,148,179],"sparsity":[45,61,128,149],"provide":[47],"the":[48,54,76,82,89,147,165,170,178],"flexibility":[49],"unstructured":[51],"pruning":[52],"with":[53,265],"runtime":[55],"efficiency":[56],"structured":[58],"approaches.":[59],"permits":[62],"arbitrarily":[63],"selecting":[64],"M":[65],"parameters":[66],"retain":[68],"from":[69],"a":[70,113,118,151,213],"contiguous":[71],"group":[72],"N":[74],"in":[75,112,231],"dense":[77],"representation.":[78],"extremely":[83],"high":[84,196],"complexity":[85],"models,":[88],"standard":[90],"sparse":[91],"fine-tuning":[92,252],"techniques":[93],"fail":[95],"generalize":[97],"well":[98],"on":[99,129,204,242,248],"downstream":[100,166,255],"tasks,":[101,218],"which":[102],"limited":[104],"data":[105],"resources.":[106],"To":[107],"address":[108],"such":[109],"an":[110],"issue":[111],"principled":[114],"manner,":[115],"we":[116,143,238,258],"introduce":[117],"new":[119],"learning":[120],"framework,":[121],"called":[122],"NxMTransformer,":[123],"induce":[125],"semi-structured":[127],"pretrained":[130],"language":[131,135],"natural":[134],"understanding":[136],"obtain":[138],"better":[139],"performance.":[140],"In":[141],"particular,":[142],"propose":[144],"formulate":[146],"as":[150],"constrained":[152],"optimization":[153],"problem":[154,181],"and":[155,219,245],"use":[156],"Alternating":[157],"Direction":[158],"Method":[159],"Multipliers":[161],"(ADMM)":[162],"optimize":[164],"tasks":[167],"while":[168,198],"taking":[169],"underlying":[171],"constraints":[173],"into":[174,182],"consideration.":[175],"ADMM":[176,250],"decomposes":[177],"sparsification":[180],"two":[183],"sub-problems":[184],"that":[185,194],"can":[186],"be":[187],"solved":[188],"sequentially,":[189],"generating":[190],"sparsified":[191],"networks":[193],"achieve":[195,226],"accuracy":[197,230,253],"being":[199],"able":[200,224],"effectively":[202],"execute":[203],"newly":[205],"released":[206],"hardware.":[207],"We":[208],"apply":[209],"our":[210,220,243],"approach":[211,244],"wide":[214],"range":[215],"NLP":[217],"proposed":[221],"method":[222],"is":[223],"1.7":[227],"points":[228],"higher":[229],"GLUE":[232],"score":[233],"than":[234],"current":[235],"practices.":[236],"Moreover,":[237],"perform":[239],"detailed":[240],"analysis":[241],"shed":[246],"light":[247],"how":[249,260],"affects":[251],"tasks.":[256],"Finally,":[257],"illustrate":[259],"NxMTransformer":[261],"achieves":[262],"performance":[263],"improvement":[264],"knowledge":[266],"distillation.":[267]},"counts_by_year":[{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":3},{"year":2023,"cited_by_count":2}],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2021-11-08T00:00:00"}