{"id":"https://openalex.org/W4412081523","doi":"https://doi.org/10.1109/tcsvt.2025.3586550","title":"Sparse-to-Dense Training: A Novel Training Scheme to Enhance Vision Transformers","display_name":"Sparse-to-Dense Training: A Novel Training Scheme to Enhance Vision Transformers","publication_year":2025,"publication_date":"2025-07-07","ids":{"openalex":"https://openalex.org/W4412081523","doi":"https://doi.org/10.1109/tcsvt.2025.3586550"},"language":"en","primary_location":{"id":"doi:10.1109/tcsvt.2025.3586550","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsvt.2025.3586550","pdf_url":null,"source":{"id":"https://openalex.org/S115173108","display_name":"IEEE Transactions on Circuits and Systems for Video Technology","issn_l":"1051-8215","issn":["1051-8215","1558-2205"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Circuits and Systems for Video Technology","raw_type":"journal-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/A5113391881","display_name":"Yongqi Huang","orcid":null},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Yongqi Huang","raw_affiliation_strings":["College of Future Information Technology, Fudan University, Shanghai, China"],"affiliations":[{"raw_affiliation_string":"College of Future Information Technology, Fudan University, Shanghai, China","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100729335","display_name":"Peng Ye","orcid":"https://orcid.org/0000-0002-5802-465X"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Peng Ye","raw_affiliation_strings":["College of Future Information Technology, Fudan University, Shanghai, China"],"affiliations":[{"raw_affiliation_string":"College of Future Information Technology, Fudan University, Shanghai, China","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114173608","display_name":"Chongjun Tu","orcid":"https://orcid.org/0009-0003-7405-2022"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chongjun Tu","raw_affiliation_strings":["College of Future Information Technology, Fudan University, Shanghai, China"],"affiliations":[{"raw_affiliation_string":"College of Future Information Technology, Fudan University, Shanghai, China","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021220108","display_name":"Tao Chen","orcid":"https://orcid.org/0000-0002-0779-9818"},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tao Chen","raw_affiliation_strings":["College of Future Information Technology, Fudan University, Shanghai, China"],"affiliations":[{"raw_affiliation_string":"College of Future Information Technology, Fudan University, Shanghai, China","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100604494","display_name":"Tong He","orcid":"https://orcid.org/0000-0003-2772-9320"},"institutions":[{"id":"https://openalex.org/I4210100255","display_name":"Beijing Academy of Artificial Intelligence","ror":"https://ror.org/016a74861","country_code":"CN","type":"other","lineage":["https://openalex.org/I4210100255"]},{"id":"https://openalex.org/I4391012619","display_name":"Shanghai Artificial Intelligence Laboratory","ror":"https://ror.org/03wkvpx79","country_code":null,"type":"facility","lineage":["https://openalex.org/I4391012619"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tong He","raw_affiliation_strings":["Shanghai Artificial Intelligence Laboratory, Shanghai, China"],"affiliations":[{"raw_affiliation_string":"Shanghai Artificial Intelligence Laboratory, Shanghai, China","institution_ids":["https://openalex.org/I4210100255","https://openalex.org/I4391012619"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5087818121","display_name":"Wanli Ouyang","orcid":"https://orcid.org/0000-0002-9163-2761"},"institutions":[{"id":"https://openalex.org/I4210100255","display_name":"Beijing Academy of Artificial Intelligence","ror":"https://ror.org/016a74861","country_code":"CN","type":"other","lineage":["https://openalex.org/I4210100255"]},{"id":"https://openalex.org/I4391012619","display_name":"Shanghai Artificial Intelligence Laboratory","ror":"https://ror.org/03wkvpx79","country_code":null,"type":"facility","lineage":["https://openalex.org/I4391012619"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wanli Ouyang","raw_affiliation_strings":["Shanghai Artificial Intelligence Laboratory, Shanghai, China"],"affiliations":[{"raw_affiliation_string":"Shanghai Artificial Intelligence Laboratory, Shanghai, China","institution_ids":["https://openalex.org/I4210100255","https://openalex.org/I4391012619"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5113391881"],"corresponding_institution_ids":["https://openalex.org/I24943067"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.21863891,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"35","issue":"12","first_page":"12329","last_page":"12340"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11408","display_name":"Advanced Optical Imaging Technologies","score":0.9948999881744385,"subfield":{"id":"https://openalex.org/subfields/2214","display_name":"Media Technology"},"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/T11408","display_name":"Advanced Optical Imaging Technologies","score":0.9948999881744385,"subfield":{"id":"https://openalex.org/subfields/2214","display_name":"Media Technology"},"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/T12153","display_name":"Advanced Optical Sensing Technologies","score":0.9860000014305115,"subfield":{"id":"https://openalex.org/subfields/3105","display_name":"Instrumentation"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.9751999974250793,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.6827404499053955},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6674736738204956},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.558295488357544},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4544163644313812},{"id":"https://openalex.org/keywords/scheme","display_name":"Scheme (mathematics)","score":0.4360746145248413},{"id":"https://openalex.org/keywords/computer-vision","display_name":"Computer vision","score":0.4186588525772095},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.4111614525318146},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3276453912258148},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.15749573707580566},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.1416662037372589},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.09398159384727478},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.09236481785774231}],"concepts":[{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.6827404499053955},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6674736738204956},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.558295488357544},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4544163644313812},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.4360746145248413},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.4186588525772095},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4111614525318146},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3276453912258148},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.15749573707580566},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.1416662037372589},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.09398159384727478},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.09236481785774231},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","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/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tcsvt.2025.3586550","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsvt.2025.3586550","pdf_url":null,"source":{"id":"https://openalex.org/S115173108","display_name":"IEEE Transactions on Circuits and Systems for Video Technology","issn_l":"1051-8215","issn":["1051-8215","1558-2205"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Circuits and Systems for Video Technology","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3977413633","display_name":null,"funder_award_id":"62071127","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5058943624","display_name":null,"funder_award_id":"2022ZD0160101","funder_id":"https://openalex.org/F4320335777","funder_display_name":"National Key Research and Development Program of China"},{"id":"https://openalex.org/G5491202952","display_name":null,"funder_award_id":"23ZR1402900","funder_id":"https://openalex.org/F4320309612","funder_display_name":"Natural Science Foundation of Shanghai"},{"id":"https://openalex.org/G5970276311","display_name":null,"funder_award_id":"2021SHZDZX0103","funder_id":"https://openalex.org/F4320335480","funder_display_name":"Guangzhou Municipal Science and Technology Project"},{"id":"https://openalex.org/G7626924423","display_name":null,"funder_award_id":"62101137","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320309612","display_name":"Natural Science Foundation of Shanghai","ror":null},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320335480","display_name":"Guangzhou Municipal Science and Technology Project","ror":null},{"id":"https://openalex.org/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":34,"referenced_works":["https://openalex.org/W1861492603","https://openalex.org/W1920022804","https://openalex.org/W2117539524","https://openalex.org/W2194775991","https://openalex.org/W2460657278","https://openalex.org/W2737258237","https://openalex.org/W2884822772","https://openalex.org/W2923014074","https://openalex.org/W2962843773","https://openalex.org/W2963150697","https://openalex.org/W2965373594","https://openalex.org/W2981609437","https://openalex.org/W3037492894","https://openalex.org/W3121523901","https://openalex.org/W3138516171","https://openalex.org/W3167976421","https://openalex.org/W3171038842","https://openalex.org/W3177096435","https://openalex.org/W3193895134","https://openalex.org/W4214636423","https://openalex.org/W4220967350","https://openalex.org/W4283815817","https://openalex.org/W4312270234","https://openalex.org/W4312349930","https://openalex.org/W4312671791","https://openalex.org/W4312897837","https://openalex.org/W4319778743","https://openalex.org/W4364382874","https://openalex.org/W4385569616","https://openalex.org/W4397026528","https://openalex.org/W4398787741","https://openalex.org/W4399485242","https://openalex.org/W4399563599","https://openalex.org/W4402671950"],"related_works":["https://openalex.org/W230091440","https://openalex.org/W2233261550","https://openalex.org/W2810751659","https://openalex.org/W258997015","https://openalex.org/W2997094352","https://openalex.org/W3216976533","https://openalex.org/W100620283","https://openalex.org/W2047028599","https://openalex.org/W4394050964","https://openalex.org/W2551249631"],"abstract_inverted_index":{"As":[0],"Vision":[1],"Transformers":[2],"(ViTs)":[3],"become":[4],"increasingly":[5],"popular":[6],"in":[7],"various":[8,157,212],"vision":[9,161,214,219],"tasks,":[10,162],"one":[11],"may":[12],"question:":[13],"<italic":[14],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[15],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">if":[16],"a":[17,44,120],"new":[18],"training":[19,30,48,54,177,184],"scheme":[20,185],"for":[21,140],"ViTs":[22,70],"exists":[23],"that":[24],"can":[25,186,202],"improve":[26,191,205],"performance":[27,193],"without":[28],"increasing":[29],"and":[31,55,82,150,159,165,170,198,217],"inference":[32,56],"computation":[33],"cost?</i>":[34],"In":[35],"this":[36,41],"paper,":[37],"we":[38,51,62,113,180],"affirmatively":[39],"answer":[40],"question":[42],"with":[43,71],"novel":[45],"Sparse-to-Dense":[46],"(S2D)":[47],"scheme.":[49,178],"Specifically,":[50],"decouple":[52],"the":[53,126,129,135,168,174,192,206],"phases":[57],"of":[58,69,173,194,208],"ViTs.":[59],"During":[60],"training,":[61,112],"replace":[63],"some":[64],"Feed-Forward":[65],"Network":[66],"(FFN)":[67],"layers":[68,106],"computationally":[72],"efficient":[73],"RUP-Mixture-of-FFN":[74],"(RUP-MoF)":[75],"layers,":[76],"each":[77,108,115],"comprising":[78],"multiple":[79],"FFN":[80,122],"experts":[81,86],"allocating":[83],"tokens":[84],"to":[85,119,134,147,190],"via":[87],"Random":[88],"Uniform":[89],"Partition":[90],"(RUP).":[91],"Furthermore,":[92],"an":[93],"additional":[94],"Experts":[95],"Weights":[96],"Averaging":[97],"(EWA)":[98],"update":[99,200],"is":[100],"performed":[101],"specifically":[102],"on":[103,211],"these":[104],"RUP-MoF":[105,116],"after":[107],"gradient":[109],"update.":[110],"After":[111],"convert":[114],"layer":[117,123],"back":[118,133],"single":[121],"by":[124],"averaging":[125],"experts,":[127],"transforming":[128],"training-time":[130],"sparse":[131],"model":[132,139],"original":[136],"dense":[137],"ViT":[138,163],"inference.":[141],"We":[142],"further":[143],"provide":[144],"theoretical":[145],"analysis":[146],"illustrate":[148],"why":[149],"how":[151],"it":[152],"works.":[153],"Comprehensive":[154],"experiments":[155],"across":[156],"2D":[158,213],"3D":[160,218],"architectures":[164],"datasets":[166,216],"validate":[167],"effectiveness":[169,207],"generalization":[171],"ability":[172],"proposed":[175],"S2D":[176,183],"Besides,":[179],"show":[181],"that,":[182],"also":[187,203],"be":[188],"applied":[189],"Transformer-based":[195],"language":[196],"models,":[197],"EWA":[199],"technique":[201],"significantly":[204],"classic":[209],"Mixture-of-Experts":[210],"small-scale":[215],"tasks.":[220]},"counts_by_year":[],"updated_date":"2026-04-09T08:11:56.329763","created_date":"2025-10-10T00:00:00"}