{"id":"https://openalex.org/W6892198691","doi":"https://doi.org/10.5075/epfl-thesis-9061","title":"Improving the Training of Compact Neural Networks for Visual Recognition","display_name":"Improving the Training of Compact Neural Networks for Visual Recognition","publication_year":2022,"publication_date":"2022-01-01","ids":{"openalex":"https://openalex.org/W6892198691","doi":"https://doi.org/10.5075/epfl-thesis-9061"},"language":"en","primary_location":{"id":"pmh:oai:infoscience.epfl.ch:298733","is_oa":true,"landing_page_url":"https://infoscience.epfl.ch/handle/20.500.14299/193059","pdf_url":null,"source":{"id":"https://openalex.org/S4306400487","display_name":"Infoscience (Ecole Polytechnique F\u00e9d\u00e9rale de Lausanne)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"doctoral thesis"},"type":"dissertation","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://infoscience.epfl.ch/handle/20.500.14299/193059","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Guo, Shuxuan","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Guo, Shuxuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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":true,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.580299973487854,"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":0.580299973487854,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.030899999663233757,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.018799999728798866,"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/convolutional-neural-network","display_name":"Convolutional neural network","score":0.7052000164985657},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.6292999982833862},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.5756999850273132},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.5677000284194946},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5558000206947327},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.4487000107765198},{"id":"https://openalex.org/keywords/cognitive-neuroscience-of-visual-object-recognition","display_name":"Cognitive neuroscience of visual object recognition","score":0.41839998960494995},{"id":"https://openalex.org/keywords/object","display_name":"Object (grammar)","score":0.40549999475479126},{"id":"https://openalex.org/keywords/linear-model","display_name":"Linear model","score":0.39259999990463257},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.39100000262260437}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7526999711990356},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.7052000164985657},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6869999766349792},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.6292999982833862},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.5756999850273132},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.5677000284194946},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5558000206947327},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4797999858856201},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.4487000107765198},{"id":"https://openalex.org/C64876066","wikidata":"https://www.wikidata.org/wiki/Q5141226","display_name":"Cognitive neuroscience of visual object recognition","level":3,"score":0.41839998960494995},{"id":"https://openalex.org/C2781238097","wikidata":"https://www.wikidata.org/wiki/Q175026","display_name":"Object (grammar)","level":2,"score":0.40549999475479126},{"id":"https://openalex.org/C163175372","wikidata":"https://www.wikidata.org/wiki/Q3339222","display_name":"Linear model","level":2,"score":0.39259999990463257},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.39100000262260437},{"id":"https://openalex.org/C49766605","wikidata":"https://www.wikidata.org/wiki/Q207643","display_name":"Linear map","level":2,"score":0.3797999918460846},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.3662000000476837},{"id":"https://openalex.org/C41045048","wikidata":"https://www.wikidata.org/wiki/Q202843","display_name":"Linear programming","level":2,"score":0.3522000014781952},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.3264999985694885},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.3012999892234802},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.30090001225471497},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.288100004196167},{"id":"https://openalex.org/C75294576","wikidata":"https://www.wikidata.org/wiki/Q5165192","display_name":"Contextual image classification","level":3,"score":0.2809000015258789},{"id":"https://openalex.org/C104122410","wikidata":"https://www.wikidata.org/wiki/Q1416406","display_name":"Network model","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C193415008","wikidata":"https://www.wikidata.org/wiki/Q639681","display_name":"Network architecture","level":2,"score":0.267300009727478},{"id":"https://openalex.org/C175202392","wikidata":"https://www.wikidata.org/wiki/Q2434543","display_name":"Time delay neural network","level":3,"score":0.2515999972820282},{"id":"https://openalex.org/C139502532","wikidata":"https://www.wikidata.org/wiki/Q1122090","display_name":"Computational intelligence","level":2,"score":0.24709999561309814},{"id":"https://openalex.org/C164752517","wikidata":"https://www.wikidata.org/wiki/Q5570875","display_name":"Global optimization","level":2,"score":0.24140000343322754},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.23340000212192535},{"id":"https://openalex.org/C2776151529","wikidata":"https://www.wikidata.org/wiki/Q3045304","display_name":"Object detection","level":3,"score":0.22349999845027924},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.21899999678134918},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.21389999985694885},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.21320000290870667},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.21070000529289246},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.20340000092983246},{"id":"https://openalex.org/C123138037","wikidata":"https://www.wikidata.org/wiki/Q5570871","display_name":"Global network","level":2,"score":0.2013999968767166},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.19850000739097595},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.19820000231266022},{"id":"https://openalex.org/C56397880","wikidata":"https://www.wikidata.org/wiki/Q6044094","display_name":"Intelligent decision support system","level":2,"score":0.19740000367164612},{"id":"https://openalex.org/C9417928","wikidata":"https://www.wikidata.org/wiki/Q1070689","display_name":"Image processing","level":3,"score":0.19110000133514404},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.186599999666214},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.18559999763965607}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:infoscience.epfl.ch:298733","is_oa":true,"landing_page_url":"https://infoscience.epfl.ch/handle/20.500.14299/193059","pdf_url":null,"source":{"id":"https://openalex.org/S4306400487","display_name":"Infoscience (Ecole Polytechnique F\u00e9d\u00e9rale de Lausanne)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"doctoral thesis"},{"id":"doi:10.5075/epfl-thesis-9061","is_oa":true,"landing_page_url":"https://doi.org/10.5075/epfl-thesis-9061","pdf_url":null,"source":null,"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Dissertation"}],"best_oa_location":{"id":"pmh:oai:infoscience.epfl.ch:298733","is_oa":true,"landing_page_url":"https://infoscience.epfl.ch/handle/20.500.14299/193059","pdf_url":null,"source":{"id":"https://openalex.org/S4306400487","display_name":"Infoscience (Ecole Polytechnique F\u00e9d\u00e9rale de Lausanne)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"doctoral thesis"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"During":[0],"the":[1,7,72,96,115,125,133,199,221,264,267,275,278,281,284,305,313,316,331,349,369,376,384,390,407,447,465,468],"Artificial":[2],"Intelligence":[3],"(AI)":[4],"revolution":[5],"of":[6,32,44,66,74,98,117,236,304,315,351,410,450,467],"past":[8],"decades,":[9],"deep":[10,27,119],"neural":[11,42],"networks":[12,43,454,470],"have":[13,18,48],"been":[14,49,359],"widely":[15],"used":[16],"and":[17,37,89,102,174,202,250,280,322,412,428],"achieved":[19],"tremendous":[20],"success":[21,116],"in":[22,58,144,383],"visual":[23,242,457],"recognition.":[24],"Unfortunately,":[25],"deploying":[26],"models":[28,371],"is":[29,69,238,258],"challenging":[30],"because":[31],"their":[33,106],"huge":[34],"model":[35,46,134,385],"size":[36,47],"computational":[38],"complexity.":[39],"Therefore,":[40],"compact":[41,77,100,161,223,434,453,469],"small":[45],"remarkably":[50],"demanded":[51],"for":[52,171,176,259,300,366,395,455],"embedded/mobile/edge":[53],"devices,":[54],"which":[55],"are":[56,136,372],"omnipresent":[57],"our":[59,327,344,422],"modern":[60],"AI":[61],"age.":[62],"The":[63,234],"main":[64],"goal":[65],"this":[67,142,145,367,439],"thesis":[68,440],"to":[70,94,104,111,114,123,158,220,274,348,380,405,442,445],"improve":[71,105,446,464],"training":[73,97,448],"\\emph{arbitrary,":[75,451],"given}":[76,452],"networks.":[78,414],"To":[79,437],"achieve":[80],"this,":[81],"we":[82,147,166,293,363],"introduce":[83,167,389],"several":[84,241,418],"methods,":[85],"including":[86,245],"linear":[87,156,183,187,195],"over-parameterization":[88,157],"two":[90,168],"novel":[91],"knowledge":[92,255,269,297,332,345,356,392],"distillation,":[93],"facilitate":[95,124],"such":[99,211],"models,":[101,120],"thus":[103],"performance.":[107,233],"Over-parameterization":[108],"was":[109],"shown":[110],"be":[112,216],"key":[113],"conventional":[118],"being":[121],"essential":[122],"optimization":[126,152,200],"during":[127,205],"training,":[128],"even":[129],"though":[130],"not":[131,333],"all":[132],"weights":[135],"necessary":[137],"at":[138,471],"inference.":[139],"Motivated":[140],"by":[141,179,263,399],"observation,":[143],"thesis,":[146],"firstly":[148],"present":[149],"a":[150,160,295,381],"general":[151],"method,":[153],"ExpandNets,":[154],"leveraging":[155],"train":[159],"network":[162,206,214,224,289],"from":[163],"scratch.":[164],"Specifically,":[165,362],"expansion":[169,196],"strategies":[170,462],"convolutional":[172],"layers":[173,178],"one":[175],"fully-connected":[177],"linearly":[180],"expanding":[181],"these":[182],"operations":[184],"into":[185],"consecutive":[186],"layers,":[188],"without":[189,225],"adding":[190],"any":[191,226],"nonlinearity.":[192],"Our":[193,253,310,415,460],"proposed":[194],"empirically":[197],"improves":[198,312],"behavior":[201],"generalization":[203],"ability":[204],"training.":[207],"At":[208],"test":[209],"time,":[210],"an":[212],"expanded":[213],"can":[215],"algebraically":[217],"contracted":[218],"back":[219],"original":[222],"information":[227],"loss,":[228],"but":[229,337],"yields":[230,429],"better":[231,426],"prediction":[232,378],"effectiveness":[235],"ExpandNets":[237],"evidenced":[239],"on":[240,319,401,417],"recognition":[243,458],"tasks,":[244],"image":[246],"classification,":[247],"object":[248,260,301],"detection,":[249,261,302],"semantic":[251],"segmentation.":[252],"first":[254,391],"distillation":[256,270,298,308,346,357,393,423],"approach":[257],"motivated":[262],"fact":[265],"that":[266,421],"recent":[268],"literature":[271],"remains":[272],"limited":[273],"scenario":[276],"where":[277,355],"student":[279,317,411,435],"teacher":[282,413],"tackle":[283],"same":[285],"task,":[286,368],"with":[287,432],"similar":[288],"architectures.":[290],"By":[291],"contrast,":[292],"propose":[294],"classifier-to-detector":[296],"method":[299,311,328,394,424],"instead":[303],"standard":[306],"detector-to-detector":[307],"strategy.":[309],"performance":[314,466],"detector":[318],"both":[320],"classification":[321],"localization.":[323],"In":[324],"other":[325],"words,":[326],"successfully":[329],"transfers":[330],"only":[334],"across":[335,339],"architectures":[336],"also":[338],"tasks.":[340,459],"We":[341,387],"then":[342],"extend":[343],"work":[347],"task":[350],"6D":[352,396],"pose":[353,397],"estimation,":[354],"has":[358],"completely":[360],"unstudied.":[361],"observe":[364],"that,":[365],"keypoint-based":[370],"less":[373],"sensitive":[374],"than":[375],"dense":[377],"ones":[379],"decrease":[382],"size.":[386],"therefore":[388],"estimation":[398],"relying":[400],"optimal":[402],"transport":[403],"theory":[404],"align":[406],"keypoint":[408],"distributions":[409],"experiments":[416],"benchmarks":[419],"show":[420],"predicts":[425],"keypoints":[427],"state-of-the-art":[430],"results":[431],"different":[433,456],"models.":[436],"summarize,":[438],"contributes":[441],"multiple":[443],"investigations":[444],"phase":[449],"diverse":[461],"consistently":[463],"inference":[472],"time.":[473]},"counts_by_year":[],"updated_date":"2026-07-23T05:56:39.545243","created_date":"2025-10-10T00:00:00"}