{"id":"https://openalex.org/W7223875352","doi":"https://doi.org/10.48550/arxiv.2610.35720","title":"Quantifying the failure of symmetry learning in linear diffusion models due to initialization and finite-sample effects","display_name":"Quantifying the failure of symmetry learning in linear diffusion models due to initialization and finite-sample effects","publication_year":2026,"publication_date":"2026-10-26","ids":{"openalex":"https://openalex.org/W7223875352","doi":"https://doi.org/10.48550/arxiv.2610.35720"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.35720","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.35720","pdf_url":"https://arxiv.org/pdf/2610.35720","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":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.35720","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5081881085","display_name":"John J. Vastola","orcid":"https://orcid.org/0000-0002-5625-2106","observed_orcids":["https://orcid.org/0000-0002-5625-2106"]},"institutions":[{"id":"https://openalex.org/I136199984","display_name":"Harvard University","ror":"https://ror.org/03vek6s52","country_code":"US","type":"education","lineage":["https://openalex.org/I136199984"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Vastola, John J.","raw_affiliation_strings":["Department of Neurobiology , Harvard Medical School Kempner"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Neurobiology , Harvard Medical School Kempner","institution_ids":["https://openalex.org/I136199984"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5159954638","display_name":"Kanaka Rajan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I136199984","display_name":"Harvard University","ror":"https://ror.org/03vek6s52","country_code":"US","type":"education","lineage":["https://openalex.org/I136199984"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Rajan, Kanaka","raw_affiliation_strings":["Department of Neurobiology , Harvard Medical School Kempner","Institute for the Study of Natural and Artificial Intelligence , Harvard University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Neurobiology , Harvard Medical School Kempner","institution_ids":["https://openalex.org/I136199984"]},{"raw_affiliation_string":"Institute for the Study of Natural and Artificial Intelligence , Harvard University","institution_ids":["https://openalex.org/I136199984"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I136199984"],"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.4043000042438507,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.4043000042438507,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.2224999964237213,"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":0.1103999987244606,"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/linear-diffusion-model","display_name":"linear diffusion model","score":0.8450000286102295},{"id":"https://openalex.org/keywords/symmetry-breaking","display_name":"symmetry breaking","score":0.9010000228881836},{"id":"https://openalex.org/keywords/initialization-bias","display_name":"initialization bias","score":0.5809999704360962},{"id":"https://openalex.org/keywords/finite-sample-effects","display_name":"finite sample effects","score":0.9409999847412109},{"id":"https://openalex.org/keywords/local-legislation","display_name":"local legislation","score":0.675000011920929},{"id":"https://openalex.org/keywords/random-matrix-theory","display_name":"random matrix theory","score":0.9959999918937683},{"id":"https://openalex.org/keywords/generative-models","display_name":"generative models","score":0.7319999933242798},{"id":"https://openalex.org/keywords/sample-efficiency","display_name":"sample efficiency","score":0.6430000066757202}],"concepts":[{"id":"https://openalex.org/C2779886137","wikidata":"https://www.wikidata.org/wiki/Q21030012","display_name":"Symmetry (geometry)","level":2,"score":0.7368999719619751},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5329999923706055},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.48339998722076416},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.4620000123977661},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.4341999888420105},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.4293000102043152},{"id":"https://openalex.org/C204795200","wikidata":"https://www.wikidata.org/wiki/Q903282","display_name":"Symmetry breaking","level":2,"score":0.3901999890804291},{"id":"https://openalex.org/C133978748","wikidata":"https://www.wikidata.org/wiki/Q15955882","display_name":"Reflection symmetry","level":2,"score":0.38999998569488525},{"id":"https://openalex.org/C32834561","wikidata":"https://www.wikidata.org/wiki/Q660730","display_name":"Subspace topology","level":2,"score":0.376800000667572},{"id":"https://openalex.org/C96469262","wikidata":"https://www.wikidata.org/wiki/Q1324364","display_name":"Homogeneous space","level":2,"score":0.35420000553131104},{"id":"https://openalex.org/C18653775","wikidata":"https://www.wikidata.org/wiki/Q1333358","display_name":"Joint probability distribution","level":2,"score":0.3517000079154968},{"id":"https://openalex.org/C12362212","wikidata":"https://www.wikidata.org/wiki/Q728435","display_name":"Linear subspace","level":2,"score":0.34299999475479126},{"id":"https://openalex.org/C114466953","wikidata":"https://www.wikidata.org/wiki/Q6034165","display_name":"Initialization","level":2,"score":0.3357999920845032},{"id":"https://openalex.org/C192423722","wikidata":"https://www.wikidata.org/wiki/Q1455736","display_name":"Global symmetry","level":4,"score":0.33489999175071716},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.32589998841285706},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.31859999895095825},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.3125999867916107},{"id":"https://openalex.org/C163175372","wikidata":"https://www.wikidata.org/wiki/Q3339222","display_name":"Linear model","level":2,"score":0.3098999857902527},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3041999936103821},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.29739999771118164},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.28780001401901245},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.28630000352859497},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.28119999170303345},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.2777999937534332},{"id":"https://openalex.org/C2777402240","wikidata":"https://www.wikidata.org/wiki/Q6783436","display_name":"Masking (illustration)","level":2,"score":0.26829999685287476},{"id":"https://openalex.org/C2779291638","wikidata":"https://www.wikidata.org/wiki/Q1455736","display_name":"Local symmetry","level":2,"score":0.26010000705718994},{"id":"https://openalex.org/C178650346","wikidata":"https://www.wikidata.org/wiki/Q201984","display_name":"Covariance","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C185142706","wikidata":"https://www.wikidata.org/wiki/Q1134404","display_name":"Covariance matrix","level":2,"score":0.2581000030040741},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2567000091075897}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.35720","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.35720","pdf_url":"https://arxiv.org/pdf/2610.35720","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":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.35720","doi":"https://doi.org/10.48550/arxiv.2610.35720","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.35720","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:2610.35720","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.35720","pdf_url":"https://arxiv.org/pdf/2610.35720","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":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G7092442049","display_name":"Integrative theory of memory and cognitive processes","funder_award_id":"220020466","funder_id":"https://openalex.org/F4320306183","funder_display_name":"James S. McDonnell Foundation"},{"id":"https://openalex.org/G833602340","display_name":"CAREER: Untangling Inter-Area Communication in the Brain Using Multi-Region Neural Networks","funder_award_id":"2046583","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320306183","display_name":"James S. McDonnell Foundation","ror":"https://ror.org/03dy4aq19"},{"id":"https://openalex.org/F4320308324","display_name":"McKnight Foundation","ror":"https://ror.org/003ghvj67"},{"id":"https://openalex.org/F4320309622","display_name":"Harvard University","ror":"https://ror.org/03vek6s52"},{"id":"https://openalex.org/F4320309949","display_name":"Canadian Institute for Advanced Research","ror":"https://ror.org/01sdtdd95"},{"id":"https://openalex.org/F4320332161","display_name":"National Institutes of Health","ror":"https://ror.org/01cwqze88"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7223875352.pdf","grobid_xml":"https://content.openalex.org/works/W7223875352.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7112803773","https://openalex.org/W4403913492","https://openalex.org/W4415796424","https://openalex.org/W4406448961","https://openalex.org/W4403344759","https://openalex.org/W4225689481","https://openalex.org/W7127616161","https://openalex.org/W4403885729","https://openalex.org/W7124729231","https://openalex.org/W7155532689"],"abstract_inverted_index":{"Data":[0],"distributions":[1],"often":[2],"possess":[3],"one":[4],"or":[5,40],"more":[6],"symmetries:":[7],"for":[8,134],"example,":[9],"all":[10,22],"orientations":[11],"of":[12,21,29,68,84,91,95],"any":[13],"specific":[14],"dog":[15,24],"are":[16,71,152],"included":[17],"in":[18,79,120,138],"the":[19,27,80,111,121],"distribution":[20,63],"possible":[23],"pictures.":[25],"Despite":[26],"omnipresence":[28],"such":[30],"symmetries,":[31,64],"modern":[32],"generative":[33],"modeling":[34],"approaches":[35],"may":[36],"fail":[37,59],"to":[38,46,60,154],"incorporate":[39],"learn":[41,61],"them,":[42],"which":[43],"can":[44],"lead":[45],"lower":[47],"sample":[48],"efficiency":[49],"and":[50,65,106,147],"poorer":[51],"generalization.":[52],"To":[53],"what":[54,66],"extent":[55],"do":[56],"diffusion":[57,86],"models":[58],"data":[62],"sources":[67,94],"symmetry":[69,96,114,136,143],"breaking":[70],"most":[72,112],"responsible?":[73],"Here,":[74],"we":[75,131],"answer":[76],"this":[77],"question":[78],"theoretically":[81],"tractable":[82],"setting":[83],"linear":[85],"models.":[87],"We":[88],"show":[89,148],"that":[90,149],"three":[92],"potential":[93],"breaking---a":[97],"non-symmetric":[98],"initial":[99],"weight":[100],"matrix,":[101],"a":[102,145],"noisy":[103],"learning":[104],"target,":[105],"finite-sample":[107],"noise---target":[108],"noise":[109],"produces":[110],"robust":[113],"breaking,":[115],"since":[116],"it":[117],"persists":[118],"even":[119],"sample-rich":[122],"regime.":[123],"Using":[124],"local":[125],"laws":[126],"from":[127],"random":[128],"matrix":[129],"theory,":[130],"derive":[132],"formulas":[133],"average":[135],"error":[137],"fairly":[139],"general":[140],"cases":[141],"involving":[142],"within":[144],"subspace":[146],"smaller":[150],"symmetries":[151],"harder":[153],"learn.":[155]},"counts_by_year":[],"updated_date":"2026-11-06T12:05:46.655746","created_date":"2026-10-28T00:00:00"}