{"id":"https://openalex.org/W3125253108","doi":"https://doi.org/10.1109/dicta52665.2021.9647143","title":"A Chaos Theory Approach to Understand Neural Network Optimization","display_name":"A Chaos Theory Approach to Understand Neural Network Optimization","publication_year":2021,"publication_date":"2021-11-01","ids":{"openalex":"https://openalex.org/W3125253108","doi":"https://doi.org/10.1109/dicta52665.2021.9647143","mag":"3125253108"},"language":"en","primary_location":{"id":"doi:10.1109/dicta52665.2021.9647143","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dicta52665.2021.9647143","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2021 Digital Image Computing: Techniques and Applications (DICTA)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5055089022","display_name":"Michele Sasdelli","orcid":"https://orcid.org/0000-0003-1021-6369"},"institutions":[{"id":"https://openalex.org/I4210127558","display_name":"Australian Centre for Robotic Vision","ror":"https://ror.org/02zv9xv82","country_code":"AU","type":"facility","lineage":["https://openalex.org/I4210127558"]},{"id":"https://openalex.org/I5681781","display_name":"The University of Adelaide","ror":"https://ror.org/00892tw58","country_code":"AU","type":"education","lineage":["https://openalex.org/I5681781"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Michele Sasdelli","raw_affiliation_strings":["Australian Institute for Machine Learning, School of Computer Science, The University of Adelaide"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Australian Institute for Machine Learning, School of Computer Science, The University of Adelaide","institution_ids":["https://openalex.org/I4210127558","https://openalex.org/I5681781"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039055995","display_name":"Thalaiyasingam Ajanthan","orcid":"https://orcid.org/0000-0002-6431-0775"},"institutions":[{"id":"https://openalex.org/I118347636","display_name":"Australian National University","ror":"https://ror.org/019wvm592","country_code":"AU","type":"education","lineage":["https://openalex.org/I118347636"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Thalaiyasingam Ajanthan","raw_affiliation_strings":["Australian National University & Amazon Australia, Work done prior to joining Amazon"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Australian National University & Amazon Australia, Work done prior to joining Amazon","institution_ids":["https://openalex.org/I118347636"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027317977","display_name":"Tat-Jun Chin","orcid":"https://orcid.org/0000-0003-2423-9342"},"institutions":[{"id":"https://openalex.org/I4210127558","display_name":"Australian Centre for Robotic Vision","ror":"https://ror.org/02zv9xv82","country_code":"AU","type":"facility","lineage":["https://openalex.org/I4210127558"]},{"id":"https://openalex.org/I5681781","display_name":"The University of Adelaide","ror":"https://ror.org/00892tw58","country_code":"AU","type":"education","lineage":["https://openalex.org/I5681781"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Tat-Jun Chin","raw_affiliation_strings":["Australian Institute for Machine Learning, School of Computer Science, The University of Adelaide"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Australian Institute for Machine Learning, School of Computer Science, The University of Adelaide","institution_ids":["https://openalex.org/I4210127558","https://openalex.org/I5681781"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029215323","display_name":"Gustavo Carneiro","orcid":"https://orcid.org/0000-0002-5571-6220"},"institutions":[{"id":"https://openalex.org/I4210127558","display_name":"Australian Centre for Robotic Vision","ror":"https://ror.org/02zv9xv82","country_code":"AU","type":"facility","lineage":["https://openalex.org/I4210127558"]},{"id":"https://openalex.org/I5681781","display_name":"The University of Adelaide","ror":"https://ror.org/00892tw58","country_code":"AU","type":"education","lineage":["https://openalex.org/I5681781"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Gustavo Carneiro","raw_affiliation_strings":["Australian Institute for Machine Learning, School of Computer Science, The University of Adelaide"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Australian Institute for Machine Learning, School of Computer Science, The University of Adelaide","institution_ids":["https://openalex.org/I4210127558","https://openalex.org/I5681781"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":"12","issue":null,"first_page":"1","last_page":"10"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9998000264167786,"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":0.9998000264167786,"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":0.9987000226974487,"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/T12611","display_name":"Neural Networks and Reservoir Computing","score":0.9975000023841858,"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/hessian-matrix","display_name":"Hessian matrix","score":0.8676219582557678},{"id":"https://openalex.org/keywords/eigenvalues-and-eigenvectors","display_name":"Eigenvalues and eigenvectors","score":0.6043391823768616},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.5661087036132812},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5511149168014526},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5356400012969971},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.4920017421245575},{"id":"https://openalex.org/keywords/range","display_name":"Range (aeronautics)","score":0.46050235629081726},{"id":"https://openalex.org/keywords/mathematical-optimization","display_name":"Mathematical optimization","score":0.448790043592453},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.4388880729675293},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.42002034187316895},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.3758136034011841},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.33665427565574646},{"id":"https://openalex.org/keywords/applied-mathematics","display_name":"Applied mathematics","score":0.32733601331710815},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.27284717559814453},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.08816704154014587}],"concepts":[{"id":"https://openalex.org/C203616005","wikidata":"https://www.wikidata.org/wiki/Q620495","display_name":"Hessian matrix","level":2,"score":0.8676219582557678},{"id":"https://openalex.org/C158693339","wikidata":"https://www.wikidata.org/wiki/Q190524","display_name":"Eigenvalues and eigenvectors","level":2,"score":0.6043391823768616},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5661087036132812},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5511149168014526},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5356400012969971},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.4920017421245575},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.46050235629081726},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.448790043592453},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.4388880729675293},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.42002034187316895},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3758136034011841},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.33665427565574646},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.32733601331710815},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.27284717559814453},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.08816704154014587},{"id":"https://openalex.org/C192562407","wikidata":"https://www.wikidata.org/wiki/Q228736","display_name":"Materials science","level":0,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C159985019","wikidata":"https://www.wikidata.org/wiki/Q181790","display_name":"Composite material","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/dicta52665.2021.9647143","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dicta52665.2021.9647143","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2021 Digital Image Computing: Techniques and Applications (DICTA)","raw_type":"proceedings-article"},{"id":"pmh:oai:digital.library.adelaide.edu.au:2440/136360","is_oa":false,"landing_page_url":"https://hdl.handle.net/2440/136360","pdf_url":null,"source":{"id":"https://openalex.org/S4306401835","display_name":"Adelaide Research & Scholarship (AR&S) (University of Adelaide)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I5681781","host_organization_name":"The University of Adelaide","host_organization_lineage":["https://openalex.org/I5681781"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"https://ieeexplore.ieee.org/xpl/conhome/9647036/proceeding","raw_type":"Conference paper"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":72,"referenced_works":["https://openalex.org/W104184427","https://openalex.org/W196761320","https://openalex.org/W1491468723","https://openalex.org/W1522301498","https://openalex.org/W1686810756","https://openalex.org/W1783671139","https://openalex.org/W1811750039","https://openalex.org/W1994616650","https://openalex.org/W2066473596","https://openalex.org/W2072998188","https://openalex.org/W2112796928","https://openalex.org/W2120420045","https://openalex.org/W2137515395","https://openalex.org/W2142426721","https://openalex.org/W2143723307","https://openalex.org/W2146502635","https://openalex.org/W2154579312","https://openalex.org/W2155894447","https://openalex.org/W2163605009","https://openalex.org/W2194775991","https://openalex.org/W2523246573","https://openalex.org/W2619516334","https://openalex.org/W2766164908","https://openalex.org/W2768267830","https://openalex.org/W2776610675","https://openalex.org/W2790586233","https://openalex.org/W2913310393","https://openalex.org/W2963177640","https://openalex.org/W2963209464","https://openalex.org/W2963317585","https://openalex.org/W2963433607","https://openalex.org/W2963587345","https://openalex.org/W2963959597","https://openalex.org/W2964072432","https://openalex.org/W2964142744","https://openalex.org/W2970247852","https://openalex.org/W2970490659","https://openalex.org/W3005560345","https://openalex.org/W3099529713","https://openalex.org/W3101544609","https://openalex.org/W3127645305","https://openalex.org/W4287758986","https://openalex.org/W4295151193","https://openalex.org/W4297779039","https://openalex.org/W4301014524","https://openalex.org/W6604254268","https://openalex.org/W6608133726","https://openalex.org/W6631190155","https://openalex.org/W6637373629","https://openalex.org/W6638214083","https://openalex.org/W6677872732","https://openalex.org/W6680610060","https://openalex.org/W6681435938","https://openalex.org/W6682423634","https://openalex.org/W6683107984","https://openalex.org/W6684191040","https://openalex.org/W6726983090","https://openalex.org/W6727249380","https://openalex.org/W6730172645","https://openalex.org/W6738491991","https://openalex.org/W6739826692","https://openalex.org/W6745410505","https://openalex.org/W6746900977","https://openalex.org/W6748278370","https://openalex.org/W6755310813","https://openalex.org/W6759121537","https://openalex.org/W6766881738","https://openalex.org/W6767185500","https://openalex.org/W6772129282","https://openalex.org/W6778944205","https://openalex.org/W6779352975","https://openalex.org/W6785119878"],"related_works":["https://openalex.org/W4312200629","https://openalex.org/W4223943233","https://openalex.org/W4360585206","https://openalex.org/W4364306694","https://openalex.org/W4309045103","https://openalex.org/W4225161397","https://openalex.org/W3014300295","https://openalex.org/W3164822677","https://openalex.org/W4250304930","https://openalex.org/W2922457425"],"abstract_inverted_index":{"Despite":[0],"the":[1,23,26,57,80,83,91,97,102,105,110,118,122,138,147,151,154,180,184,188,196,219],"complicated":[2],"structure":[3],"of":[4,28,82,86,101,104,112,121,133,146,153,172,183,198,221],"modern":[5],"deep":[6,46,222],"neural":[7,47],"network":[8,48],"architectures,":[9],"they":[10],"are":[11,232],"still":[12],"optimized":[13],"with":[14,143],"algorithms":[15],"based":[16],"on":[17,195,240],"Stochastic":[18],"Gradient":[19],"Descent":[20],"(SGD).":[21],"However,":[22],"reason":[24],"behind":[25],"effectiveness":[27,197],"SGD":[29,70,87,136,177,199],"is":[30],"not":[31],"well":[32],"understood,":[33],"making":[34],"its":[35,72],"study":[36,62],"an":[37,113,214],"active":[38],"research":[39],"area.":[40],"In":[41,74],"this":[42],"paper,":[43],"we":[44,76,125],"formulate":[45],"optimization":[49,220],"as":[50],"a":[51,130,169,208],"dynamical":[52],"system":[53],"and":[54,71,174,230,237],"show":[55,127],"that":[56,79,128,176,211,227],"rigorous":[58],"theory":[59],"developed":[60],"to":[61,68,96,116,150,166],"chaotic":[63],"systems":[64],"can":[65,163],"be":[66,164,167],"useful":[67],"understand":[69],"variants.":[73],"particular,":[75],"first":[77],"observe":[78],"inverse":[81,152],"instability":[84],"timescale":[85],"optimization,":[88],"represented":[89],"by":[90],"largest":[92,119,144,181],"Lyapunov":[93],"exponent,":[94],"corresponds":[95],"most":[98],"negative":[99],"eigenvalue":[100,120,145,182],"Hessian":[103,148,185],"loss.":[106],"This":[107,157,191],"observation":[108],"enables":[109],"introduction":[111],"efficient":[114],"method":[115,210],"estimate":[117],"Hessian.":[123],"Then,":[124],"empirically":[126],"for":[129,218],"large":[131,170],"range":[132,171],"learning":[134,155,161,216,223],"rates,":[135],"traverses":[137],"loss":[139,189,238],"landscape":[140],"across":[141,234],"regions":[142],"similar":[149],"rate.":[156],"explains":[158],"why":[159],"effective":[160],"rates":[162],"found":[165],"within":[168],"values":[173],"shows":[175],"implicitly":[178],"uses":[179],"while":[186],"traversing":[187],"landscape.":[190],"sheds":[192],"some":[193],"light":[194],"over":[200],"more":[201],"sophisticated":[202],"second-order":[203],"methods.":[204],"We":[205,225],"also":[206],"propose":[207],"quasi-Newton":[209],"dynamically":[212],"estimates":[213],"optimal":[215],"rate":[217],"models.":[224],"demonstrate":[226],"our":[228],"observations":[229],"methods":[231],"robust":[233],"different":[235],"architectures":[236],"functions":[239],"CIFAR-10":[241],"dataset.":[242]},"counts_by_year":[{"year":2024,"cited_by_count":1},{"year":2021,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}