{"id":"https://openalex.org/W3213489681","doi":null,"title":"Quasi-potential theory for escape problem: Quantitative sharpness effect on SGD's escape from local minima.","display_name":"Quasi-potential theory for escape problem: Quantitative sharpness effect on SGD's escape from local minima.","publication_year":2021,"publication_date":"2021-11-07","ids":{"openalex":"https://openalex.org/W3213489681","mag":"3213489681"},"language":"en","primary_location":{"id":"mag:3213489681","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2111.04004","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2111.04004","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5074221409","display_name":"Hikaru Ibayashi","orcid":"https://orcid.org/0000-0001-6983-1927","observed_orcids":["https://orcid.org/0000-0001-6983-1927"]},"institutions":[{"id":"https://openalex.org/I1174212","display_name":"University of Southern California","ror":"https://ror.org/03taz7m60","country_code":"US","type":"education","lineage":["https://openalex.org/I1174212"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hikaru Ibayashi","raw_affiliation_strings":["University  of Southern California"],"raw_orcid":"https://orcid.org/0000-0001-6983-1927","affiliations":[{"raw_affiliation_string":"University  of Southern California","institution_ids":["https://openalex.org/I1174212"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5066308746","display_name":"Masaaki Imaizumi","orcid":"https://orcid.org/0000-0001-6186-613X","observed_orcids":["https://orcid.org/0000-0001-6186-613X"]},"institutions":[{"id":"https://openalex.org/I74801974","display_name":"The University of Tokyo","ror":"https://ror.org/057zh3y96","country_code":"JP","type":"education","lineage":["https://openalex.org/I74801974"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Masaaki Imaizumi","raw_affiliation_strings":["Univ.\\ of Tokyo"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Univ.\\ of Tokyo","institution_ids":["https://openalex.org/I74801974"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"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":false,"primary_topic":{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9990000128746033,"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":0.9990000128746033,"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/T10320","display_name":"Neural Networks and Applications","score":9.999999747378752e-05,"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/T12535","display_name":"Machine Learning and Data Classification","score":9.999999747378752e-05,"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/stochastic-gradient-descent","display_name":"stochastic gradient descent","score":0.8579999804496765},{"id":"https://openalex.org/keywords/nonconvex-optimization","display_name":"non-convex optimization","score":0.8650000095367432},{"id":"https://openalex.org/keywords/stochastic-dynamical-systems","display_name":"stochastic dynamical systems","score":0.8529999852180481},{"id":"https://openalex.org/keywords/generalization","display_name":"generalization","score":0.7390000224113464},{"id":"https://openalex.org/keywords/neural-network","display_name":"artificial neural networks","score":0.5649999976158142}],"concepts":[{"id":"https://openalex.org/C186633575","wikidata":"https://www.wikidata.org/wiki/Q845060","display_name":"Maxima and minima","level":2,"score":0.8262301087379456},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.6069974899291992},{"id":"https://openalex.org/C112680207","wikidata":"https://www.wikidata.org/wiki/Q714886","display_name":"Regular polygon","level":2,"score":0.5598602294921875},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5322840809822083},{"id":"https://openalex.org/C178650346","wikidata":"https://www.wikidata.org/wiki/Q201984","display_name":"Covariance","level":2,"score":0.48467662930488586},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.45929473638534546},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.45025599002838135},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.42387160658836365},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.35776233673095703},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.333471417427063},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.31517457962036133},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2727934718132019},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.21490392088890076},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.14002037048339844},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.1312844157218933},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.07203283905982971},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.06811407208442688}],"mesh":[],"locations_count":1,"locations":[{"id":"mag:3213489681","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2111.04004","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null}],"best_oa_location":{"id":"mag:3213489681","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2111.04004","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":55,"referenced_works":["https://openalex.org/W196871588","https://openalex.org/W1521738998","https://openalex.org/W1522301498","https://openalex.org/W1843889522","https://openalex.org/W1965154800","https://openalex.org/W1999844375","https://openalex.org/W2049631384","https://openalex.org/W2059120410","https://openalex.org/W2143163787","https://openalex.org/W2163605009","https://openalex.org/W2523060838","https://openalex.org/W2552194003","https://openalex.org/W2605372163","https://openalex.org/W2626325961","https://openalex.org/W2660589364","https://openalex.org/W2731468224","https://openalex.org/W2738410460","https://openalex.org/W2763894180","https://openalex.org/W2765861987","https://openalex.org/W2766164908","https://openalex.org/W2766371994","https://openalex.org/W2768267830","https://openalex.org/W2788923859","https://openalex.org/W2790867397","https://openalex.org/W2793333878","https://openalex.org/W2796146910","https://openalex.org/W2798707604","https://openalex.org/W2799042347","https://openalex.org/W2804386825","https://openalex.org/W2891688105","https://openalex.org/W2896721680","https://openalex.org/W2904142648","https://openalex.org/W2911183980","https://openalex.org/W2911682203","https://openalex.org/W2911853241","https://openalex.org/W2922220370","https://openalex.org/W2947518183","https://openalex.org/W2951412272","https://openalex.org/W2952230511","https://openalex.org/W2962839956","https://openalex.org/W2962932339","https://openalex.org/W2962933129","https://openalex.org/W2963069632","https://openalex.org/W2963384892","https://openalex.org/W2963655672","https://openalex.org/W2964286272","https://openalex.org/W2970490659","https://openalex.org/W2981382427","https://openalex.org/W2990730264","https://openalex.org/W2992043355","https://openalex.org/W3035443457","https://openalex.org/W3040620603","https://openalex.org/W3117548433","https://openalex.org/W3118431319","https://openalex.org/W3122542623"],"related_works":["https://openalex.org/W4294828903","https://openalex.org/W3004700632","https://openalex.org/W3177888558","https://openalex.org/W4286793238","https://openalex.org/W2911183980","https://openalex.org/W3118431319","https://openalex.org/W3037239953","https://openalex.org/W2804386825","https://openalex.org/W3186933740","https://openalex.org/W3160865277"],"abstract_inverted_index":{"We":[0,105,172],"develop":[1,89],"a":[2,10,90,99,120,127],"quantitative":[3],"theory":[4,92,100,110,180],"on":[5,25],"an":[6,53,69,79,156],"escape":[7,66,95,154],"problem":[8,67],"of":[9,20,22,47,101,116,123,147,155],"stochastic":[11,102],"gradient":[12,124],"descent":[13],"(SGD)":[14],"algorithm":[15],"and":[16,119,158,167],"investigate":[17],"the":[18,26,48,94,108,145,152,160,165,170],"effect":[19,166],"sharpness":[21,146],"loss":[23,63,117,148],"surfaces":[24,118,149],"escape.":[27,171],"Deep":[28],"learning":[29],"has":[30,39],"achieved":[31],"tremendous":[32],"success":[33],"in":[34,126,136],"various":[35,42],"domains,":[36],"however,":[37],"it":[38],"opened":[40],"up":[41],"theoretical":[43,140],"open":[44],"questions.":[45],"One":[46],"typical":[49],"questions":[50],"is":[51,68],"why":[52],"SGD":[54,80],"can":[55,111],"find":[56],"parameters":[57],"that":[58,107,143],"generalize":[59],"well":[60],"over":[61],"non-convex":[62],"surfaces.":[64],"An":[65],"approach":[70],"to":[71,151,176],"tackle":[72],"this":[73,86],"question,":[74],"which":[75],"investigates":[76],"how":[77],"efficiently":[78],"escapes":[81],"from":[82],"local":[83],"minima.":[84],"In":[85],"paper,":[87],"we":[88],"quasi-potential":[91,109],"for":[93],"problem,":[96],"by":[97],"applying":[98],"dynamical":[103],"systems.":[104],"show":[106],"handle":[112],"both":[113],"geometric":[114],"properties":[115],"covariance":[121],"structure":[122,163],"noise":[125,162],"unified":[128],"manner,":[129],"while":[130],"they":[131],"have":[132],"been":[133],"separately":[134],"studied":[135],"previous":[137],"works.":[138],"Our":[139],"results":[141],"imply":[142],"(i)":[144],"contributes":[150],"slow":[153],"SGD,":[157],"(ii)":[159],"SGD's":[161],"cancels":[164],"exponentially":[168],"accelerates":[169],"also":[173],"conduct":[174],"experiments":[175],"empirically":[177],"validate":[178],"our":[179],"using":[181],"neural":[182],"networks":[183],"trained":[184],"with":[185],"real":[186],"data.":[187]},"counts_by_year":[],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2021-11-22T00:00:00"}