{"id":"https://openalex.org/W4385484715","doi":"https://doi.org/10.1109/ijcnn54540.2023.10191206","title":"Gradient Sparsification For Masked Fine-Tuning of Transformers","display_name":"Gradient Sparsification For Masked Fine-Tuning of Transformers","publication_year":2023,"publication_date":"2023-06-18","ids":{"openalex":"https://openalex.org/W4385484715","doi":"https://doi.org/10.1109/ijcnn54540.2023.10191206"},"language":"en","primary_location":{"id":"doi:10.1109/ijcnn54540.2023.10191206","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn54540.2023.10191206","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 International Joint Conference on Neural Networks (IJCNN)","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/A5083742715","display_name":"James O\u2019Neill","orcid":"https://orcid.org/0000-0002-5196-7719"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"James O'Neill","raw_affiliation_strings":["Huawei Ireland Research Center,Dublin,Ireland","Huawei Ireland Research Center, Dublin, Ireland"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Ireland Research Center,Dublin,Ireland","institution_ids":[]},{"raw_affiliation_string":"Huawei Ireland Research Center, Dublin, Ireland","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5053023517","display_name":"Sourav Dutta","orcid":"https://orcid.org/0000-0002-8934-9166"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sourav Dutta","raw_affiliation_strings":["Huawei Ireland Research Center,Dublin,Ireland","Huawei Ireland Research Center, Dublin, Ireland"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Ireland Research Center,Dublin,Ireland","institution_ids":[]},{"raw_affiliation_string":"Huawei Ireland Research Center, Dublin, Ireland","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.3263,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.64310749,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":95},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"8"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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":0.9998999834060669,"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":0.9990000128746033,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.9969000220298767,"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/computer-science","display_name":"Computer science","score":0.7924046516418457},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.598842442035675},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.5897996425628662},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.5477924346923828},{"id":"https://openalex.org/keywords/fine-tuning","display_name":"Fine-tuning","score":0.49536073207855225},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.48242053389549255},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.47227391600608826},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.45359936356544495},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3472822308540344},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.11587628722190857}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7924046516418457},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.598842442035675},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5897996425628662},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.5477924346923828},{"id":"https://openalex.org/C157524613","wikidata":"https://www.wikidata.org/wiki/Q2828883","display_name":"Fine-tuning","level":2,"score":0.49536073207855225},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.48242053389549255},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.47227391600608826},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.45359936356544495},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3472822308540344},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.11587628722190857},{"id":"https://openalex.org/C205649164","wikidata":"https://www.wikidata.org/wiki/Q1071","display_name":"Geography","level":0,"score":0.0},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","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/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C13280743","wikidata":"https://www.wikidata.org/wiki/Q131089","display_name":"Geodesy","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:10.1109/ijcnn54540.2023.10191206","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn54540.2023.10191206","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.5600000023841858,"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":44,"referenced_works":["https://openalex.org/W2095705004","https://openalex.org/W2896457183","https://openalex.org/W2914120296","https://openalex.org/W2952087486","https://openalex.org/W2952436057","https://openalex.org/W2963026768","https://openalex.org/W2963250244","https://openalex.org/W2964303773","https://openalex.org/W2970193165","https://openalex.org/W2970352191","https://openalex.org/W2970752815","https://openalex.org/W2971031524","https://openalex.org/W2971207485","https://openalex.org/W2975381464","https://openalex.org/W3013840636","https://openalex.org/W3018647120","https://openalex.org/W3020268419","https://openalex.org/W3035390927","https://openalex.org/W3035497479","https://openalex.org/W3085479580","https://openalex.org/W3093298238","https://openalex.org/W3094491777","https://openalex.org/W3099793224","https://openalex.org/W3101498587","https://openalex.org/W3102483398","https://openalex.org/W3103334733","https://openalex.org/W3122044994","https://openalex.org/W3126961594","https://openalex.org/W3153675281","https://openalex.org/W3156326119","https://openalex.org/W4206178588","https://openalex.org/W6674330103","https://openalex.org/W6755207826","https://openalex.org/W6759455113","https://openalex.org/W6759579507","https://openalex.org/W6764288440","https://openalex.org/W6764398373","https://openalex.org/W6768080748","https://openalex.org/W6774623936","https://openalex.org/W6775935039","https://openalex.org/W6777155081","https://openalex.org/W6782454744","https://openalex.org/W6783566936","https://openalex.org/W6790064496"],"related_works":["https://openalex.org/W3037551068","https://openalex.org/W3049463507","https://openalex.org/W2936497627","https://openalex.org/W4288365749","https://openalex.org/W3013624417","https://openalex.org/W4287826556","https://openalex.org/W3023285645","https://openalex.org/W4287598411","https://openalex.org/W3098382480","https://openalex.org/W3198458223"],"abstract_inverted_index":{"Fine-tuning":[0,14],"pretrained":[1,23,112],"self-supervised":[2],"language":[3,113],"models":[4,114],"is":[5,78,87,144,163],"widely":[6],"adopted":[7],"for":[8,115,172],"transfer":[9],"learning":[10],"to":[11,66,90,106,110],"downstream":[12],"tasks.":[13],"can":[15],"be":[16],"achieved":[17],"by":[18,36,51],"freezing":[19],"gradients":[20,28,54,109,134],"of":[21,29,55,93,128],"the":[22,49,136,154],"network":[24],"and":[25,70,123,146,175,179],"only":[26],"updating":[27],"a":[30,46,126],"newly":[31],"added":[32],"classification":[33],"layer,":[34],"or":[35],"performing":[37],"gradient":[38,129,141],"updates":[39],"on":[40,153],"all":[41],"parameters.":[42],"Gradual":[43],"unfreezing":[44,53,83,95],"makes":[45],"trade-off":[47,67],"between":[48,68],"two":[50],"gradually":[52,82],"whole":[56],"layers":[57,84],"during":[58,135],"training.":[59],"This":[60],"has":[61],"been":[62],"an":[63],"effective":[64],"strategy":[65],"storage":[69],"training":[71,86],"speed":[72],"with":[73,158,189],"generalization":[74],"performance.":[75,100,119],"However,":[76],"it":[77,191],"not":[79,193],"clear":[80],"whether":[81],"throughout":[85],"optimal,":[88],"compared":[89],"sparse":[91,145],"variants":[92,124],"gradual":[94,149,180],"which":[96],"may":[97],"improve":[98],"fine-tuning":[99,178],"In":[101],"this":[102],"paper,":[103],"we":[104],"propose":[105],"stochastically":[107],"mask":[108,133],"regularize":[111],"improving":[116],"overall":[117],"fine-tuned":[118],"We":[120],"introduce":[121],"GradDrop":[122,143,162,186],"thereof,":[125],"class":[127],"sparsification":[130],"methods":[131,166],"that":[132,161,167],"backward":[137],"pass,":[138],"acting":[139],"as":[140,197],"noise.":[142],"stochastic":[147],"unlike":[148],"freezing.":[150],"Extensive":[151],"experiments":[152],"multilingual":[155],"XGLUE":[156],"benchmark":[157],"XLMR-Large":[159],"show":[160],"competitive":[164],"against":[165],"use":[168],"additional":[169],"translated":[170],"data":[171],"intermediate":[173],"pretraining":[174],"outperforms":[176],"standard":[177],"unfreezing.":[181],"A":[182],"post-analysis":[183],"shows":[184],"how":[185],"improves":[187],"performance":[188],"languages":[190],"was":[192],"trained":[194],"on,":[195],"such":[196],"under-resourced":[198],"languages.":[199]},"counts_by_year":[{"year":2025,"cited_by_count":1},{"year":2023,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}