{"id":"https://openalex.org/W4415020702","doi":"https://doi.org/10.48550/arxiv.2505.13900","title":"New Evidence of the Two-Phase Learning Dynamics of Neural Networks","display_name":"New Evidence of the Two-Phase Learning Dynamics of Neural Networks","publication_year":2025,"publication_date":"2025-05-20","ids":{"openalex":"https://openalex.org/W4415020702","doi":"https://doi.org/10.48550/arxiv.2505.13900"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2505.13900","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.13900","pdf_url":"https://arxiv.org/pdf/2505.13900","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2505.13900","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102878822","display_name":"Zhanpeng Zhou","orcid":"https://orcid.org/0000-0003-4065-2524","observed_orcids":["https://orcid.org/0000-0003-4065-2524"]},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhou, Zhanpeng","raw_affiliation_strings":["Shanghai Jiao Tong University ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University ,","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048510347","display_name":"Yongyi Yang","orcid":"https://orcid.org/0000-0003-3564-3260","observed_orcids":["https://orcid.org/0000-0003-3564-3260","https://orcid.org/0009-0005-5551-9875"]},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yang, Yongyi","raw_affiliation_strings":["University of Michigan ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Michigan ,","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066053285","display_name":"Mahito Sugiyama","orcid":"https://orcid.org/0000-0001-5907-9831","observed_orcids":["https://orcid.org/0000-0001-5907-9831"]},"institutions":[{"id":"https://openalex.org/I184597095","display_name":"National Institute of Informatics","ror":"https://ror.org/04ksd4g47","country_code":"JP","type":"facility","lineage":["https://openalex.org/I1319490839","https://openalex.org/I184597095","https://openalex.org/I4210158934"]},{"id":"https://openalex.org/I200475212","display_name":"The Graduate University for Advanced Studies, SOKENDAI","ror":"https://ror.org/0516ah480","country_code":"JP","type":"education","lineage":["https://openalex.org/I200475212"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Sugiyama, Mahito","raw_affiliation_strings":["National Institute of Informatics ,","The Graduate University for Advanced Studies , SOKENDAI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National Institute of Informatics ,","institution_ids":["https://openalex.org/I184597095"]},{"raw_affiliation_string":"The Graduate University for Advanced Studies , SOKENDAI","institution_ids":["https://openalex.org/I200475212"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5087158377","display_name":"Junchi Yan","orcid":"https://orcid.org/0000-0001-9639-7679","observed_orcids":["https://orcid.org/0000-0001-9639-7679"]},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yan, Junchi","raw_affiliation_strings":["Shanghai Jiao Tong University ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University ,","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10320","display_name":"Neural Networks and Applications","score":0.49950000643730164,"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/T10320","display_name":"Neural Networks and Applications","score":0.49950000643730164,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.43549999594688416,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.021800000220537186,"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/training-dynamics","display_name":"training dynamics","score":0.6990000009536743},{"id":"https://openalex.org/keywords/phase-transitions","display_name":"phase transitions","score":0.5910000205039978},{"id":"https://openalex.org/keywords/cone-effect","display_name":"cone effect","score":0.7620000243186951},{"id":"https://openalex.org/keywords/parameter-perturbation","display_name":"parameter perturbation","score":0.7260000109672546},{"id":"https://openalex.org/keywords/sensitivity-to-initial-conditions","display_name":"sensitivity to initial conditions","score":0.7279999852180481},{"id":"https://openalex.org/keywords/deep-neural-network","display_name":"deep neural network","score":0.5540000200271606}],"concepts":[{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5875999927520752},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5738999843597412},{"id":"https://openalex.org/C177918212","wikidata":"https://www.wikidata.org/wiki/Q803623","display_name":"Perturbation (astronomy)","level":2,"score":0.5705000162124634},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5608999729156494},{"id":"https://openalex.org/C2777052490","wikidata":"https://www.wikidata.org/wiki/Q5072826","display_name":"Chaotic","level":2,"score":0.54830002784729},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.4564000070095062},{"id":"https://openalex.org/C194232998","wikidata":"https://www.wikidata.org/wiki/Q1606712","display_name":"Transition (genetics)","level":3,"score":0.429500013589859},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.3919999897480011},{"id":"https://openalex.org/C149288129","wikidata":"https://www.wikidata.org/wiki/Q185357","display_name":"Phase transition","level":2,"score":0.3831000030040741},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.3783999979496002},{"id":"https://openalex.org/C138187205","wikidata":"https://www.wikidata.org/wiki/Q131251","display_name":"Tangent","level":2,"score":0.36959999799728394},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.3467000126838684},{"id":"https://openalex.org/C147168706","wikidata":"https://www.wikidata.org/wiki/Q1457734","display_name":"Recurrent neural network","level":3,"score":0.3361000120639801},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.3285999894142151},{"id":"https://openalex.org/C145912823","wikidata":"https://www.wikidata.org/wiki/Q113558","display_name":"Dynamics (music)","level":2,"score":0.3249000012874603},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.30090001225471497},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.29089999198913574},{"id":"https://openalex.org/C2775936607","wikidata":"https://www.wikidata.org/wiki/Q466845","display_name":"Tracking (education)","level":2,"score":0.28760001063346863},{"id":"https://openalex.org/C123757187","wikidata":"https://www.wikidata.org/wiki/Q9195957","display_name":"Network dynamics","level":2,"score":0.28040000796318054},{"id":"https://openalex.org/C79379906","wikidata":"https://www.wikidata.org/wiki/Q3174497","display_name":"Dynamical systems theory","level":2,"score":0.2597000002861023},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.25540000200271606}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2505.13900","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.13900","pdf_url":"https://arxiv.org/pdf/2505.13900","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2505.13900","doi":"https://doi.org/10.48550/arxiv.2505.13900","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2505.13900","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2505.13900","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.13900","pdf_url":"https://arxiv.org/pdf/2505.13900","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415020702.pdf","grobid_xml":"https://content.openalex.org/works/W4415020702.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4415150931","https://openalex.org/W7133184177","https://openalex.org/W3101581426","https://openalex.org/W3123178201","https://openalex.org/W3009686669","https://openalex.org/W6910755541","https://openalex.org/W4388444893","https://openalex.org/W4387839287","https://openalex.org/W2963504252","https://openalex.org/W4397021729"],"abstract_inverted_index":{"Understanding":[0],"how":[1,168],"deep":[2,63,169],"neural":[3],"networks":[4,170],"learn":[5],"remains":[6],"a":[7,24,50,91,141,155,163],"fundamental":[8],"challenge":[9],"in":[10],"modern":[11],"machine":[12],"learning.":[13,64],"A":[14],"growing":[15],"body":[16],"of":[17,31,62,84,120,167],"evidence":[18],"suggests":[19],"that":[20,45,57,81,129],"training":[21],"dynamics":[22],"undergo":[23],"distinct":[25],"phase":[26],"transition,":[27],"yet":[28],"our":[29],"understanding":[30],"this":[32,38,131],"transition":[33,92,132,171],"is":[34,99,107,138],"still":[35],"incomplete.":[36],"In":[37],"paper,":[39],"we":[40,79,127],"introduce":[41],"an":[42,71,100],"interval-wise":[43],"perspective":[44],"compares":[46],"network":[47,86,106],"states":[48],"across":[49],"time":[51],"window,":[52],"revealing":[53],"two":[54],"new":[55],"phenomena":[56],"illuminate":[58],"the":[59,82,85,88,105,118,121,134,146],"two-phase":[60],"nature":[61],"i)":[65],"\\textbf{The":[66,114],"Chaos":[67],"Effect.}":[68,116],"By":[69],"injecting":[70],"imperceptibly":[72],"small":[73],"parameter":[74],"perturbation":[75,89],"at":[76],"various":[77],"stages,":[78],"show":[80],"response":[83],"to":[87,95,110,140,149,175],"exhibits":[90],"from":[93,172],"chaotic":[94],"stable,":[96],"suggesting":[97],"there":[98],"early":[101],"critical":[102],"period":[103],"where":[104],"highly":[108],"sensitive":[109,173],"initial":[111],"conditions;":[112],"ii)":[113],"Cone":[115],"Tracking":[117],"evolution":[119],"empirical":[122],"Neural":[123],"Tangent":[124],"Kernel":[125],"(eNTK),":[126],"find":[128],"after":[130],"point":[133],"model's":[135],"functional":[136],"trajectory":[137],"confined":[139],"narrow":[142],"cone-shaped":[143],"subset:":[144],"while":[145],"kernel":[147],"continues":[148],"change,":[150],"it":[151],"gets":[152],"trapped":[153],"into":[154],"tight":[156],"angular":[157],"region.":[158],"Together,":[159],"these":[160],"effects":[161],"provide":[162],"structural,":[164],"dynamical":[165],"view":[166],"exploration":[174],"stable":[176],"refinement":[177],"during":[178],"training.":[179]},"counts_by_year":[],"updated_date":"2026-09-30T09:27:27.532063","created_date":"2025-10-10T00:00:00"}