{"id":"https://openalex.org/W2966523297","doi":"https://doi.org/10.24963/ijcai.2019/578","title":"BN-invariant Sharpness Regularizes the Training Model to Better Generalization","display_name":"BN-invariant Sharpness Regularizes the Training Model to Better Generalization","publication_year":2019,"publication_date":"2019-07-28","ids":{"openalex":"https://openalex.org/W2966523297","doi":"https://doi.org/10.24963/ijcai.2019/578","mag":"2966523297"},"language":"en","primary_location":{"id":"doi:10.24963/ijcai.2019/578","is_oa":true,"landing_page_url":"https://doi.org/10.24963/ijcai.2019/578","pdf_url":"https://www.ijcai.org/proceedings/2019/0578.pdf","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://www.ijcai.org/proceedings/2019/0578.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5087695030","display_name":"Mingyang Yi","orcid":"https://orcid.org/0009-0009-0638-2306"},"institutions":[{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]},{"id":"https://openalex.org/I4210120485","display_name":"Academy of Mathematics and Systems Science","ror":"https://ror.org/02jkmyk67","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210120485"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Mingyang Yi","raw_affiliation_strings":["Academy of Mathematics and Systems Science","University of Chinese Academy of Sciences"],"affiliations":[{"raw_affiliation_string":"Academy of Mathematics and Systems Science","institution_ids":["https://openalex.org/I4210120485"]},{"raw_affiliation_string":"University of Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080701389","display_name":"Huishuai Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Huishuai Zhang","raw_affiliation_strings":["Microsoft Research"],"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100344481","display_name":"Wei Chen","orcid":"https://orcid.org/0000-0002-7663-278X"},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Wei Chen","raw_affiliation_strings":["Microsoft Research"],"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078600050","display_name":"Zhi-Ming Ma","orcid":"https://orcid.org/0000-0001-6453-4368"},"institutions":[{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]},{"id":"https://openalex.org/I4210120485","display_name":"Academy of Mathematics and Systems Science","ror":"https://ror.org/02jkmyk67","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210120485"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhi-Ming Ma","raw_affiliation_strings":["Academy of Mathematics and Systems Science","University of Chinese Academy of Sciences"],"affiliations":[{"raw_affiliation_string":"Academy of Mathematics and Systems Science","institution_ids":["https://openalex.org/I4210120485"]},{"raw_affiliation_string":"University of Chinese Academy of Sciences","institution_ids":["https://openalex.org/I4210165038"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101884287","display_name":"Tie\u2010Yan Liu","orcid":"https://orcid.org/0000-0002-0476-8020"},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Tie-Yan Liu","raw_affiliation_strings":["Microsoft Research"],"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5087695030"],"corresponding_institution_ids":["https://openalex.org/I4210120485","https://openalex.org/I4210165038"],"apc_list":null,"apc_paid":null,"fwci":0.2024,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":{"value":0.53271356,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":94},"biblio":{"volume":null,"issue":null,"first_page":"4164","last_page":"4170"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9987999796867371,"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":0.9987999796867371,"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":0.9980999827384949,"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/T10812","display_name":"Human Pose and Action Recognition","score":0.996999979019165,"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/maxima-and-minima","display_name":"Maxima and minima","score":0.8511993288993835},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.7153286337852478},{"id":"https://openalex.org/keywords/normalization","display_name":"Normalization (sociology)","score":0.566682755947113},{"id":"https://openalex.org/keywords/invariant","display_name":"Invariant (physics)","score":0.5538220405578613},{"id":"https://openalex.org/keywords/scale-invariance","display_name":"Scale invariance","score":0.5168124437332153},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.509522020816803},{"id":"https://openalex.org/keywords/maxima","display_name":"Maxima","score":0.5035447478294373},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.49694326519966125},{"id":"https://openalex.org/keywords/ball","display_name":"Ball (mathematics)","score":0.475038081407547},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.4737442433834076},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.4693905711174011},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.45878392457962036},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3991660177707672},{"id":"https://openalex.org/keywords/mathematical-analysis","display_name":"Mathematical analysis","score":0.2445821762084961}],"concepts":[{"id":"https://openalex.org/C186633575","wikidata":"https://www.wikidata.org/wiki/Q845060","display_name":"Maxima and minima","level":2,"score":0.8511993288993835},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.7153286337852478},{"id":"https://openalex.org/C136886441","wikidata":"https://www.wikidata.org/wiki/Q926129","display_name":"Normalization (sociology)","level":2,"score":0.566682755947113},{"id":"https://openalex.org/C190470478","wikidata":"https://www.wikidata.org/wiki/Q2370229","display_name":"Invariant (physics)","level":2,"score":0.5538220405578613},{"id":"https://openalex.org/C135593079","wikidata":"https://www.wikidata.org/wiki/Q1750766","display_name":"Scale invariance","level":2,"score":0.5168124437332153},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.509522020816803},{"id":"https://openalex.org/C91528185","wikidata":"https://www.wikidata.org/wiki/Q520952","display_name":"Maxima","level":3,"score":0.5035447478294373},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.49694326519966125},{"id":"https://openalex.org/C122041747","wikidata":"https://www.wikidata.org/wiki/Q838611","display_name":"Ball (mathematics)","level":2,"score":0.475038081407547},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4737442433834076},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4693905711174011},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.45878392457962036},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3991660177707672},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.2445821762084961},{"id":"https://openalex.org/C52119013","wikidata":"https://www.wikidata.org/wiki/Q50637","display_name":"Art history","level":1,"score":0.0},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.0},{"id":"https://openalex.org/C144024400","wikidata":"https://www.wikidata.org/wiki/Q21201","display_name":"Sociology","level":0,"score":0.0},{"id":"https://openalex.org/C37914503","wikidata":"https://www.wikidata.org/wiki/Q156495","display_name":"Mathematical physics","level":1,"score":0.0},{"id":"https://openalex.org/C19165224","wikidata":"https://www.wikidata.org/wiki/Q23404","display_name":"Anthropology","level":1,"score":0.0},{"id":"https://openalex.org/C142362112","wikidata":"https://www.wikidata.org/wiki/Q735","display_name":"Art","level":0,"score":0.0},{"id":"https://openalex.org/C554144382","wikidata":"https://www.wikidata.org/wiki/Q213156","display_name":"Performance art","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.24963/ijcai.2019/578","is_oa":true,"landing_page_url":"https://doi.org/10.24963/ijcai.2019/578","pdf_url":"https://www.ijcai.org/proceedings/2019/0578.pdf","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.24963/ijcai.2019/578","is_oa":true,"landing_page_url":"https://doi.org/10.24963/ijcai.2019/578","pdf_url":"https://www.ijcai.org/proceedings/2019/0578.pdf","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W2966523297.pdf","grobid_xml":"https://content.openalex.org/works/W2966523297.grobid-xml"},"referenced_works_count":35,"referenced_works":["https://openalex.org/W1607832978","https://openalex.org/W1686810756","https://openalex.org/W1836465849","https://openalex.org/W1899249567","https://openalex.org/W2014384147","https://openalex.org/W2029029543","https://openalex.org/W2112796928","https://openalex.org/W2151034540","https://openalex.org/W2163605009","https://openalex.org/W2402588523","https://openalex.org/W2523060838","https://openalex.org/W2552194003","https://openalex.org/W2605372163","https://openalex.org/W2617242334","https://openalex.org/W2622379787","https://openalex.org/W2732724430","https://openalex.org/W2750854376","https://openalex.org/W2759764766","https://openalex.org/W2777662428","https://openalex.org/W2891160178","https://openalex.org/W2891872827","https://openalex.org/W2899748887","https://openalex.org/W2899790086","https://openalex.org/W2904142648","https://openalex.org/W2950220847","https://openalex.org/W2950928354","https://openalex.org/W2963959597","https://openalex.org/W2966623823","https://openalex.org/W3137695714","https://openalex.org/W4205293427","https://openalex.org/W4299971819","https://openalex.org/W6791858558","https://openalex.org/W6803771590","https://openalex.org/W6863631769","https://openalex.org/W6864014924"],"related_works":["https://openalex.org/W2050191802","https://openalex.org/W3105193347","https://openalex.org/W1973781123","https://openalex.org/W2160559052","https://openalex.org/W181473267","https://openalex.org/W4232457976","https://openalex.org/W1835819466","https://openalex.org/W2156081102","https://openalex.org/W2797691919","https://openalex.org/W2350245891"],"abstract_inverted_index":{"It":[0,75],"is":[1],"arguably":[2],"believed":[3],"that":[4,16],"flatter":[5],"minima":[6],"can":[7],"generalize":[8],"better.":[9],"However,":[10],"it":[11],"has":[12],"been":[13],"pointed":[14],"out":[15],"the":[17,25,28,77,84,88,100,108,114,118,126],"usual":[18],"definitions":[19],"of":[20,30,36,63,79,90],"sharpness,":[21,64],"which":[22,66],"consider":[23],"either":[24],"maxima":[26],"or":[27],"integral":[29,85,106],"loss":[31],"over":[32,139],"a":[33,61,95],"delta":[34],"ball":[35],"parameters":[37],"around":[38],"minima,":[39],"cannot":[40],"give":[41],"consistent":[42,68],"measurement":[43],"for":[44,70],"scale":[45,80,89],"invariant":[46],"neural":[47],"networks,":[48],"e.g.,":[49],"networks":[50,72],"with":[51,87],"batch":[52],"normalization":[53],"layer.":[54],"In":[55],"this":[56],"paper,":[57],"we":[58,93,112],"first":[59],"propose":[60],"measure":[62],"BN-Sharpness,":[65],"gives":[67],"value":[69],"equivalent":[71],"under":[73],"BN.":[74],"achieves":[76,132],"property":[78],"invariance":[81],"by":[82],"connecting":[83],"diameter":[86],"parameter.":[91],"Then":[92],"present":[94],"computation-efficient":[96],"way":[97],"to":[98,116,124],"calculate":[99],"BN-sharpness":[101,115],"approximately":[102],"i.e.,":[103],"one":[104],"dimensional":[105],"along":[107],"\"sharpest\"":[109],"direction.":[110],"Furthermore,":[111],"use":[113],"regularize":[117],"training":[119],"and":[120],"design":[121],"an":[122],"algorithm":[123,131],"minimize":[125],"new":[127],"regularized":[128],"objective.":[129],"Our":[130],"considerably":[133],"better":[134],"performance":[135],"than":[136],"vanilla":[137],"SGD":[138],"various":[140],"experiment":[141],"settings.":[142]},"counts_by_year":[{"year":2022,"cited_by_count":1},{"year":2021,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}