{"id":"https://openalex.org/W7154562945","doi":"https://doi.org/10.14428/esann/2026.es2026-117","title":"Where to grow: a surprisingly straightforward criterion to detect under-expressive layers","display_name":"Where to grow: a surprisingly straightforward criterion to detect under-expressive layers","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7154562945","doi":"https://doi.org/10.14428/esann/2026.es2026-117"},"language":null,"primary_location":{"id":"doi:10.14428/esann/2026.es2026-117","is_oa":true,"landing_page_url":"https://doi.org/10.14428/esann/2026.es2026-117","pdf_url":"https://doi.org/10.14428/esann/2026.es2026-117","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ESANN 2026 proceesdings","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.14428/esann/2026.es2026-117","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133733719","display_name":"Yifan Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yifan WANG","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113968179","display_name":"Julien Mille","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Julien MILLE","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5083547309","display_name":"Moncef Hidane","orcid":"https://orcid.org/0000-0003-2175-4365"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Moncef HIDANE","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.42916804,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"715","last_page":"720"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11737","display_name":"Advanced Materials and Mechanics","score":0.03790000081062317,"subfield":{"id":"https://openalex.org/subfields/2210","display_name":"Mechanical 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/T11737","display_name":"Advanced Materials and Mechanics","score":0.03790000081062317,"subfield":{"id":"https://openalex.org/subfields/2210","display_name":"Mechanical 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/T10184","display_name":"Plant Molecular Biology Research","score":0.03449999913573265,"subfield":{"id":"https://openalex.org/subfields/1110","display_name":"Plant Science"},"field":{"id":"https://openalex.org/fields/11","display_name":"Agricultural and Biological Sciences"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T11650","display_name":"Skin and Cellular Biology Research","score":0.02879999950528145,"subfield":{"id":"https://openalex.org/subfields/1307","display_name":"Cell Biology"},"field":{"id":"https://openalex.org/fields/13","display_name":"Biochemistry, Genetics and Molecular Biology"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.3393999934196472},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.2777999937534332},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.2578999996185303},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.25119999051094055},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.2475000023841858}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.37720000743865967},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.37220001220703125},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.3393999934196472},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.326200008392334},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.28130000829696655},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.2777999937534332},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.2578999996185303},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.25119999051094055},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.2475000023841858},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.23890000581741333}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.14428/esann/2026.es2026-117","is_oa":true,"landing_page_url":"https://doi.org/10.14428/esann/2026.es2026-117","pdf_url":"https://doi.org/10.14428/esann/2026.es2026-117","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ESANN 2026 proceesdings","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.14428/esann/2026.es2026-117","is_oa":true,"landing_page_url":"https://doi.org/10.14428/esann/2026.es2026-117","pdf_url":"https://doi.org/10.14428/esann/2026.es2026-117","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ESANN 2026 proceesdings","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/13","score":0.42624562978744507,"display_name":"Climate action"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7154562945.pdf","grobid_xml":"https://content.openalex.org/works/W7154562945.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"idea":[1],"of":[2,7,298],"gradually":[3,131],"increasing":[4],"the":[5,35,40,43,159,182,267,290,295,310],"capacity":[6,242],"a":[8,49,105,115,120,126,212,257,270,330],"neural":[9,100],"network,":[10],"during":[11,134],"or":[12,146,220,249,301],"after":[13],"training,":[14],"has":[15],"gained":[16],"attention":[17],"in":[18],"recent":[19],"years.Online":[20],"network":[21,195],"growing":[22,99],"raises":[23],"three":[24],"fundamental":[25],"questions:":[26],"when,":[27],"where,":[28],"and":[29,90,92,130,165,223,239,279,319,322],"how":[30,166,240],"to":[31,52,107,113,143,154,157,162,169,177,193,216,237,243,303],"expand.Among":[32],"all":[33],"questions,":[34],"present":[36],"paper":[37],"focuses":[38],"on":[39,60,247],"where.We":[41],"introduce":[42,309],"Normalized":[44],"Activation":[45],"Gradient":[46],"Norm":[47],"(NAGN),":[48],"lightweight":[50],"criterion":[51],"detect":[53],"under-expressive":[54],"layers":[55,145,180,221],"using":[56],"standard":[57],"backpropagation":[58],"signals.Experiments":[59],"image":[61],"classification":[62],"demonstrate":[63],"that":[64,71],"this":[65,262],"approach":[66],"consistently":[67],"discovers":[68],"compact":[69],"architectures":[70],"match":[72],"larger":[73],"static":[74],"baselines":[75],"with":[76],"reduced":[77],"training":[78,135,191],"costs.":[79],"Growing":[80],"networksTraditional":[81],"deep":[82],"learning":[83],"models":[84],"often":[85],"require":[86],"large":[87],"computational":[88,110],"resources":[89],"datasets,":[91],"they":[93],"usually":[94],"contain":[95],"many":[96,167],"redundant":[97],"parameters.Online":[98],"networks":[101],"were":[102],"introduced":[103],"as":[104],"way":[106],"lower":[108],"these":[109],"demands":[111],"and/or":[112],"find":[114],"more":[116],"efficient":[117],"architecture":[118,129],"for":[119,252],"given":[121],"task.These":[122],"methods":[123,173,197,259],"start":[124],"from":[125],"small":[127,213],"seed":[128],"expand":[132,238],"it":[133],"[1,":[136],"2,":[137],"3],":[138],"either":[139],"by":[140,147,325],"adding":[141,175,327],"units":[142,176],"existing":[144],"introducing":[148],"new":[149,203],"layers.A":[150],"key":[151],"challenge":[152],"is":[153],"decide":[155],"when":[156,201,227,236],"trigger":[158,304],"growth,":[160],"where":[161,253],"apply":[163],"it,":[164],"units/layers":[168],"add":[170],"[2].Early":[171],"online":[172],"keep":[174],"pre-specified":[178],"hidden":[179],"until":[181],"model":[183],"reaches":[184],"some":[185,224],"performance":[186],"target":[187,261],"[4,5].Later":[188],"work":[189],"uses":[190],"signals":[192],"guide":[194],"growth.Some":[196],"maximize":[198],"gradient":[199],"norms":[200],"initializing":[202],"neurons":[204,219,328],"under":[205],"fixed":[206],"growth":[207,254],"schedules":[208],"[6],":[209],"others":[210],"search":[211],"functional":[214,318],"neighborhood":[215],"select":[217],"additional":[218],"[7],":[222],"increase":[225],"depth":[226],"validation":[228],"accuracy":[229],"stops":[230],"improving":[231],"[8].These":[232],"approaches":[233],"mainly":[234],"address":[235],"much":[241],"add,":[244],"but":[245],"rely":[246],"hand-designed":[248],"restricted":[250],"choices":[251],"can":[255],"occur.Only":[256],"few":[258],"directly":[260],"question.Liu":[263],"et":[264,286,306],"al.[1]":[265],"use":[266],"splitting":[268],"matrix,":[269],"truncated":[271],"Hessian":[272],"whose":[273],"minimum":[274],"eigenvalues":[275],"identify":[276],"which":[277,281,293,313],"neurons,":[278],"therefore":[280],"layers,":[282],"should":[283],"be":[284],"expanded.Maile":[285],"al.":[287,307],"[2]":[288],"propose":[289],"NORTH*":[291],"framework,":[292],"monitors":[294],"numerical":[296],"rank":[297],"post-activation":[299],"outputs":[300],"weights":[302],"growth.Verbockhaven":[305],"[3]":[308],"TINY":[311],"algorithm,":[312],"measures":[314],"expressivity":[315],"gaps":[316],"between":[317],"realizable":[320],"gradients":[321],"resolves":[323],"them":[324],"optimally":[326],"through":[329],"quadratic":[331],"optimization":[332],"step.":[333]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-04-17T00:00:00"}