{"id":"https://openalex.org/W7204193781","doi":"https://doi.org/10.48550/arxiv.2608.21561","title":"Loss-Parameterized Fisher Width Along Learning Trajectories","display_name":"Loss-Parameterized Fisher Width Along Learning Trajectories","publication_year":2026,"publication_date":"2026-08-21","ids":{"openalex":"https://openalex.org/W7204193781","doi":"https://doi.org/10.48550/arxiv.2608.21561"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.21561","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21561","pdf_url":"https://arxiv.org/pdf/2608.21561","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/2608.21561","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5016120331","display_name":"Vu Khac Ky","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ky, Vu Khac","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5016120331"],"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9814000129699707,"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":0.9814000129699707,"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":0.007000000216066837,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.003000000026077032,"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/gaussian-width","display_name":"Gaussian width","score":0.6399999856948853},{"id":"https://openalex.org/keywords/learning-trajectories","display_name":"learning trajectories","score":0.7099999785423279},{"id":"https://openalex.org/keywords/teacher-student-models","display_name":"teacher-student models","score":0.6069999933242798},{"id":"https://openalex.org/keywords/stability-bounds","display_name":"stability bounds","score":0.4320000112056732},{"id":"https://openalex.org/keywords/adam-optimizer","display_name":"Adam optimizer","score":0.36399999260902405},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"gradient descent","score":0.4880000054836273},{"id":"https://openalex.org/keywords/stochastic-gradient-descent","display_name":"stochastic gradient descent","score":0.925000011920929},{"id":"https://openalex.org/keywords/neural-network-training-dynamics","display_name":"neural network training dynamics","score":0.4880000054836273}],"concepts":[{"id":"https://openalex.org/C29406490","wikidata":"https://www.wikidata.org/wiki/Q1420659","display_name":"Fisher information","level":2,"score":0.7437000274658203},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.7383999824523926},{"id":"https://openalex.org/C2908647359","wikidata":"https://www.wikidata.org/wiki/Q2625603","display_name":"Population","level":2,"score":0.5777999758720398},{"id":"https://openalex.org/C130367717","wikidata":"https://www.wikidata.org/wiki/Q189791","display_name":"Diagonal","level":2,"score":0.47540000081062317},{"id":"https://openalex.org/C158622935","wikidata":"https://www.wikidata.org/wiki/Q660848","display_name":"Nonlinear system","level":2,"score":0.4325999915599823},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.41609999537467957},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.40880000591278076},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.40700000524520874},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.4059999883174896},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.3896999955177307},{"id":"https://openalex.org/C83282275","wikidata":"https://www.wikidata.org/wiki/Q7435350","display_name":"Scoring algorithm","level":2,"score":0.34769999980926514},{"id":"https://openalex.org/C61031149","wikidata":"https://www.wikidata.org/wiki/Q1153200","display_name":"Fisher equation","level":4,"score":0.34450000524520874},{"id":"https://openalex.org/C62354387","wikidata":"https://www.wikidata.org/wiki/Q875399","display_name":"Boundary (topology)","level":2,"score":0.34220001101493835},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.3271999955177307},{"id":"https://openalex.org/C202887219","wikidata":"https://www.wikidata.org/wiki/Q3895221","display_name":"Parametrization (atmospheric modeling)","level":3,"score":0.32670000195503235},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.30489999055862427},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.3046000003814697},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.299699991941452},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.296999990940094},{"id":"https://openalex.org/C41949839","wikidata":"https://www.wikidata.org/wiki/Q1756677","display_name":"Stability theory","level":3,"score":0.28060001134872437},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2761000096797943},{"id":"https://openalex.org/C187834632","wikidata":"https://www.wikidata.org/wiki/Q188804","display_name":"Factorization","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C190470478","wikidata":"https://www.wikidata.org/wiki/Q2370229","display_name":"Invariant (physics)","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C2778067643","wikidata":"https://www.wikidata.org/wiki/Q166507","display_name":"Interval (graph theory)","level":2,"score":0.26170000433921814},{"id":"https://openalex.org/C146324458","wikidata":"https://www.wikidata.org/wiki/Q1202673","display_name":"Uniform norm","level":2,"score":0.26089999079704285},{"id":"https://openalex.org/C163175372","wikidata":"https://www.wikidata.org/wiki/Q3339222","display_name":"Linear model","level":2,"score":0.2549000084400177}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.21561","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21561","pdf_url":"https://arxiv.org/pdf/2608.21561","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.2608.21561","doi":"https://doi.org/10.48550/arxiv.2608.21561","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.21561","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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.21561","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21561","pdf_url":"https://arxiv.org/pdf/2608.21561","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":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204193781.pdf","grobid_xml":"https://content.openalex.org/works/W7204193781.grobid-xml"},"referenced_works_count":11,"referenced_works":["https://openalex.org/W74946117","https://openalex.org/W1573082642","https://openalex.org/W1970789124","https://openalex.org/W1995842804","https://openalex.org/W2025223969","https://openalex.org/W2167077875","https://openalex.org/W2963095610","https://openalex.org/W3009686669","https://openalex.org/W3065974826","https://openalex.org/W7165110421","https://openalex.org/W7170809197"],"related_works":["https://openalex.org/W7170809197","https://openalex.org/W7165110421","https://openalex.org/W3117548433","https://openalex.org/W4399694537","https://openalex.org/W7133190362","https://openalex.org/W7214986236","https://openalex.org/W7165355408","https://openalex.org/W7133200361","https://openalex.org/W7220703587","https://openalex.org/W4312057594"],"abstract_inverted_index":{"Fisher":[0,15,80,84,172],"width":[1,5],"measures":[2],"the":[3,13,60,101,121,125,152],"Gaussian":[4],"of":[6,171],"a":[7,46,55,129,133,148,164],"probe":[8],"set":[9],"after":[10],"deformation":[11],"by":[12],"local":[14],"geometry.":[16],"We":[17,38,86],"study":[18],"its":[19],"evolution":[20],"along":[21],"learning":[22],"trajectories":[23],"and":[24,45,77,82,103,124,138],"ask":[25],"when":[26],"training":[27],"loss":[28,67,169],"can":[29],"serve":[30],"as":[31],"an":[32,41],"effective":[33],"coordinate":[34],"for":[35,50,100,108],"this":[36,95],"quantity.":[37],"first":[39],"derive":[40],"exact":[42],"trace--shape":[43],"factorization":[44],"deterministic":[47],"stability":[48],"bound":[49],"fixed":[51,153],"compact":[52],"probes.":[53],"In":[54,128],"population":[56,90,126],"Gaussian-teacher":[57],"logistic":[58],"model,":[59],"teacher-aligned":[61],"state":[62],"is":[63],"extremal":[64],"on":[65],"every":[66],"level":[68],"below":[69],"$\\log":[70],"2$:":[71],"it":[72],"has":[73],"minimal":[74],"parameter":[75],"norm":[76],"maximizes":[78],"both":[79],"trace":[81],"Euclidean-ball":[83],"width.":[85,173],"then":[87],"show":[88],"that":[89],"gradient":[91],"flow":[92],"asymptotically":[93],"selects":[94],"branch,":[96],"with":[97,132],"explicit":[98],"rates":[99],"aligned":[102],"orthogonal":[104],"coordinates.":[105],"This":[106],"yields,":[107],"$d\\geq2$,":[109],"\\[":[110],"\\frac{w_F(B_2^d;\u03b8(t))}":[111],"{\\sqrt{L(\u03b8(t))}}":[112],"\\longrightarrow":[113],"\\frac{\\sqrt6}\u03c0\\mathbb":[114],"E[\u03c7_{d-1}].":[115],"\\]":[116],"Controlled":[117],"full-Fisher":[118],"experiments":[119],"support":[120,163],"matched-loss":[122],"branch":[123],"predictions.":[127],"nonlinear":[130],"MLP":[131],"diagonal":[134],"model-Fisher":[135],"approximation,":[136],"GD":[137],"SGD":[139],"remain":[140],"close":[141],"at":[142],"matched":[143],"loss,":[144],"whereas":[145],"Adam":[146],"follows":[147],"substantially":[149],"displaced":[150],"branch;":[151],"probes":[154],"tested":[155],"retain":[156],"highly":[157],"similar":[158],"temporal":[159],"shapes.":[160],"These":[161],"results":[162],"branchwise,":[165],"rather":[166],"than":[167],"universal,":[168],"parametrization":[170]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-26T00:00:00"}