{"id":"https://openalex.org/W7137908851","doi":"https://doi.org/10.1609/aaai.v40i24.39074","title":"Unpacking the Implicit Norm Dynamics of Sharpness-Aware Minimization in Tensorized Models","display_name":"Unpacking the Implicit Norm Dynamics of Sharpness-Aware Minimization in Tensorized Models","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7137908851","doi":"https://doi.org/10.1609/aaai.v40i24.39074"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i24.39074","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i24.39074","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/39074/43036","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://ojs.aaai.org/index.php/AAAI/article/download/39074/43036","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5105872324","display_name":"Tianxiao Cao","orcid":null},"institutions":[{"id":"https://openalex.org/I22299242","display_name":"Kyoto University","ror":"https://ror.org/02kpeqv85","country_code":"JP","type":"education","lineage":["https://openalex.org/I22299242"]}],"countries":["JP"],"is_corresponding":true,"raw_author_name":"Tianxiao Cao","raw_affiliation_strings":["Kyoto University"],"affiliations":[{"raw_affiliation_string":"Kyoto University","institution_ids":["https://openalex.org/I22299242"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024780466","display_name":"Kyohei Atarashi","orcid":null},"institutions":[{"id":"https://openalex.org/I22299242","display_name":"Kyoto University","ror":"https://ror.org/02kpeqv85","country_code":"JP","type":"education","lineage":["https://openalex.org/I22299242"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Kyohei Atarashi","raw_affiliation_strings":["Kyoto University"],"affiliations":[{"raw_affiliation_string":"Kyoto University","institution_ids":["https://openalex.org/I22299242"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5129735651","display_name":"Hisashi Kashima","orcid":null},"institutions":[{"id":"https://openalex.org/I22299242","display_name":"Kyoto University","ror":"https://ror.org/02kpeqv85","country_code":"JP","type":"education","lineage":["https://openalex.org/I22299242"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Hisashi Kashima","raw_affiliation_strings":["Kyoto University"],"affiliations":[{"raw_affiliation_string":"Kyoto University","institution_ids":["https://openalex.org/I22299242"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5105872324"],"corresponding_institution_ids":["https://openalex.org/I22299242"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.12980529,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"24","first_page":"19907","last_page":"19915"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.357699990272522,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.357699990272522,"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/T12303","display_name":"Tensor decomposition and applications","score":0.28610000014305115,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.08370000123977661,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/norm","display_name":"Norm (philosophy)","score":0.5924999713897705},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.4903999865055084},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.478300005197525},{"id":"https://openalex.org/keywords/balanced-flow","display_name":"Balanced flow","score":0.4546000063419342},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.4410000145435333},{"id":"https://openalex.org/keywords/minification","display_name":"Minification","score":0.4237000048160553},{"id":"https://openalex.org/keywords/simple","display_name":"Simple (philosophy)","score":0.3619999885559082}],"concepts":[{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.5924999713897705},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5645999908447266},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.4903999865055084},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.478300005197525},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.47049999237060547},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.4578999876976013},{"id":"https://openalex.org/C167879884","wikidata":"https://www.wikidata.org/wiki/Q727568","display_name":"Balanced flow","level":2,"score":0.4546000063419342},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.4410000145435333},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.4237000048160553},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.3619999885559082},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.3321000039577484},{"id":"https://openalex.org/C2777256151","wikidata":"https://www.wikidata.org/wiki/Q7897273","display_name":"Unpacking","level":2,"score":0.3273000121116638},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3206999897956848},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.3122999966144562},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3052000105381012},{"id":"https://openalex.org/C178650346","wikidata":"https://www.wikidata.org/wiki/Q201984","display_name":"Covariance","level":2,"score":0.2937000095844269},{"id":"https://openalex.org/C97542219","wikidata":"https://www.wikidata.org/wiki/Q497863","display_name":"SIMPLE algorithm","level":2,"score":0.2777999937534332},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.2775999903678894},{"id":"https://openalex.org/C92207270","wikidata":"https://www.wikidata.org/wiki/Q939253","display_name":"Matrix norm","level":3,"score":0.2750000059604645}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i24.39074","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i24.39074","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/39074/43036","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i24.39074","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i24.39074","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/39074/43036","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320334789","display_name":"Japan Science and Technology Agency","ror":"https://ror.org/00097mb19"},{"id":"https://openalex.org/F4320338075","display_name":"Core Research for Evolutional Science and Technology","ror":"https://ror.org/00097mb19"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7137908851.pdf","grobid_xml":"https://content.openalex.org/works/W7137908851.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Sharpness-Aware":[0],"Minimization":[1],"(SAM)":[2],"has":[3],"been":[4],"proven":[5],"to":[6,50],"be":[7],"an":[8],"effective":[9,116],"optimization":[10],"technique":[11],"for":[12],"improving":[13],"generalization":[14],"in":[15,28,35,57,131],"overparameterized":[16],"models.":[17,60],"While":[18],"prior":[19],"works":[20],"have":[21],"explored":[22],"the":[23,52,63,98],"implicit":[24,90],"regularization":[25,125],"of":[26,55,65,72,92],"SAM":[27,56,81],"simple":[29,114],"two-core":[30],"scale-invariant":[31,40],"settings,":[32],"its":[33,78],"behavior":[34,126],"more":[36],"general":[37,58],"tensorized":[38,59],"or":[39,152],"models":[41],"remains":[42],"underexplored.":[43],"In":[44],"this":[45,124],"work,":[46],"we":[47,111],"leverage":[48],"scale-invariance":[49],"analyze":[51],"norm":[53,74],"dynamics":[54],"We":[61,86],"introduce":[62],"notion":[64],"Norm":[66,93],"Deviation":[67,94],"as":[68],"a":[69,113,132],"global":[70],"measure":[71],"core":[73,101,129],"imbalance,":[75],"and":[76,103,144],"derive":[77],"evolution":[79],"under":[80],"using":[82],"gradient":[83,105],"flow":[84],"analysis.":[85],"show":[87],"that":[88,148],"SAM's":[89],"control":[91],"is":[95],"governed":[96],"by":[97,108,127],"covariance":[99],"between":[100],"norms":[102,130],"their":[104],"magnitudes.":[106],"Motivated":[107],"these":[109],"findings,":[110],"propose":[112],"yet":[115],"method,":[117],"Deviation-Aware":[118],"Scaling":[119],"(DAS),":[120],"which":[121],"explicitly":[122],"mimics":[123],"scaling":[128],"data-adaptive":[133],"manner.":[134],"Our":[135],"experiments":[136],"across":[137],"tensor":[138],"completion,":[139],"noisy":[140],"training,":[141],"model":[142],"compression,":[143],"parameter-efficient":[145],"fine-tuning":[146],"confirm":[147],"DAS":[149],"achieves":[150],"competitive":[151],"improved":[153],"performance":[154],"over":[155],"SAM,":[156],"while":[157],"offering":[158],"reduced":[159],"computational":[160],"overhead.":[161]},"counts_by_year":[],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2026-03-18T00:00:00"}