{"id":"https://openalex.org/W4387946345","doi":"https://doi.org/10.48550/arxiv.2310.14901","title":"Series of Hessian-Vector Products for Tractable Saddle-Free Newton Optimisation of Neural Networks","display_name":"Series of Hessian-Vector Products for Tractable Saddle-Free Newton Optimisation of Neural Networks","publication_year":2023,"publication_date":"2023-10-23","ids":{"openalex":"https://openalex.org/W4387946345","doi":"https://doi.org/10.48550/arxiv.2310.14901"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2310.14901","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.14901","pdf_url":"https://arxiv.org/pdf/2310.14901","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2310.14901","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5082464744","display_name":"Elre T. Oldewage","orcid":"https://orcid.org/0000-0002-0568-8700","observed_orcids":["https://orcid.org/0000-0002-0568-8700"]},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Oldewage, Elre T.","raw_affiliation_strings":["Department of Engineering University of Cambridge"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Engineering University of Cambridge","institution_ids":["https://openalex.org/I241749"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039277324","display_name":"Ross M. Clarke","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Clarke, Ross M.","raw_affiliation_strings":["Department of Engineering University of Cambridge"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Engineering University of Cambridge","institution_ids":["https://openalex.org/I241749"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5102890597","display_name":"Jos\u00e9 Miguel Hern\u00e1ndez-Lobato","orcid":"https://orcid.org/0000-0001-7610-949X","observed_orcids":["https://orcid.org/0000-0001-7610-949X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hern\u00e1ndez-Lobato, Jos\u00e9 Miguel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9970999956130981,"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":0.9970999956130981,"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/T10963","display_name":"Advanced Optimization Algorithms Research","score":0.001500000013038516,"subfield":{"id":"https://openalex.org/subfields/2612","display_name":"Numerical Analysis"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10320","display_name":"Neural Networks and Applications","score":0.0003000000142492354,"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-vector-product","display_name":"Hessian-vector product","score":0.8220000267028809},{"id":"https://openalex.org/keywords/quasi-newton-methods","display_name":"quasi-Newton methods","score":0.8529999852180481},{"id":"https://openalex.org/keywords/neural-network-training","display_name":"neural network training","score":0.871999979019165},{"id":"https://openalex.org/keywords/nonconvex-optimization","display_name":"non-convex optimization","score":0.871999979019165},{"id":"https://openalex.org/keywords/gradient-preconditioning","display_name":"gradient preconditioning","score":0.7649999856948853},{"id":"https://openalex.org/keywords/resnet-18","display_name":"ResNet-18","score":0.8560000061988831}],"concepts":[{"id":"https://openalex.org/C203616005","wikidata":"https://www.wikidata.org/wiki/Q620495","display_name":"Hessian matrix","level":2,"score":0.8345586657524109},{"id":"https://openalex.org/C143724316","wikidata":"https://www.wikidata.org/wiki/Q312468","display_name":"Series (stratigraphy)","level":2,"score":0.6022128462791443},{"id":"https://openalex.org/C2681867","wikidata":"https://www.wikidata.org/wiki/Q690935","display_name":"Saddle point","level":2,"score":0.5695763230323792},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5377059578895569},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4068235456943512},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.38426706194877625},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.36705249547958374},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3316325545310974},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.23646679520606995},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.05137285590171814},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0},{"id":"https://openalex.org/C151730666","wikidata":"https://www.wikidata.org/wiki/Q7205","display_name":"Paleontology","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2310.14901","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.14901","pdf_url":"https://arxiv.org/pdf/2310.14901","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.2310.14901","doi":"https://doi.org/10.48550/arxiv.2310.14901","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2310.14901","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:2310.14901","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.14901","pdf_url":"https://arxiv.org/pdf/2310.14901","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":[{"id":"https://openalex.org/G2465893628","display_name":"Peta-5: A National Facility for Petascale Data Intensive Computation and Analytics","funder_award_id":"EP/P020259/1","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"},{"id":"https://openalex.org/G358276009","display_name":null,"funder_award_id":"EP/P020259","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"},{"id":"https://openalex.org/G3887492883","display_name":null,"funder_award_id":"2107369","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"},{"id":"https://openalex.org/G8402139738","display_name":null,"funder_award_id":"EP/S515334/1","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"}],"funders":[{"id":"https://openalex.org/F4320312927","display_name":"Dell EMC","ror":"https://ror.org/05rejmm18"},{"id":"https://openalex.org/F4320334627","display_name":"Engineering and Physical Sciences Research Council","ror":"https://ror.org/0439y7842"},{"id":"https://openalex.org/F4320334632","display_name":"Science and Technology Facilities Council","ror":"https://ror.org/057g20z61"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4387946345.pdf","grobid_xml":"https://content.openalex.org/works/W4387946345.grobid-xml"},"referenced_works_count":8,"referenced_works":["https://openalex.org/W32532385","https://openalex.org/W2020688938","https://openalex.org/W2324838733","https://openalex.org/W2561193401","https://openalex.org/W2750384547","https://openalex.org/W3193576386","https://openalex.org/W4292122331","https://openalex.org/W4384662421"],"related_works":["https://openalex.org/W2245520501","https://openalex.org/W3188189585","https://openalex.org/W2410732425","https://openalex.org/W2777827707","https://openalex.org/W3180893216","https://openalex.org/W3080080747","https://openalex.org/W6995687949","https://openalex.org/W4307204268","https://openalex.org/W2531810927","https://openalex.org/W7128536709"],"abstract_inverted_index":{"Despite":[0],"their":[1],"popularity":[2],"in":[3,16,45,134,143],"the":[4,20,32,41,64,72,82,92,109],"field":[5],"of":[6,57,113,146],"continuous":[7],"optimisation,":[8],"second-order":[9,131],"quasi-Newton":[10],"methods":[11,133],"are":[12],"challenging":[13],"to":[14,34,61,69,98,127],"apply":[15],"machine":[17],"learning,":[18],"as":[19,44,84],"Hessian":[21,75],"matrix":[22],"is":[23,29,123],"intractably":[24],"large.":[25],"This":[26],"computational":[27],"burden":[28],"exacerbated":[30],"by":[31,39],"need":[33],"address":[35],"non-convexity,":[36],"for":[37],"instance":[38],"modifying":[40],"Hessian's":[42],"eigenvalues":[43],"Saddle-Free":[46],"Newton":[47],"methods.":[48],"We":[49,140],"propose":[50],"an":[51],"optimisation":[52,67,120,132,138],"algorithm":[53,68,121],"which":[54,87,122],"addresses":[55],"both":[56,135],"these":[58],"concerns":[59],"-":[60],"our":[62],"knowledge,":[63],"first":[65],"efficiently-scalable":[66],"asymptotically":[70],"use":[71],"exact":[73],"inverse":[74],"with":[76],"absolute-value":[77],"eigenvalues.":[78],"Our":[79],"method":[80],"frames":[81],"problem":[83],"a":[85,100,118,144,149],"series":[86,116],"principally":[88],"square-roots":[89],"and":[90,125,130,137],"inverts":[91],"squared":[93],"Hessian,":[94],"then":[95],"uses":[96],"it":[97],"precondition":[99],"gradient":[101],"vector,":[102],"all":[103],"without":[104],"explicitly":[105],"computing":[106],"or":[107],"eigendecomposing":[108],"Hessian.":[110],"A":[111],"truncation":[112],"this":[114,142],"infinite":[115],"provides":[117],"new":[119],"scalable":[124],"comparable":[126],"other":[128],"first-":[129],"runtime":[136],"performance.":[139],"demonstrate":[141],"variety":[145],"settings,":[147],"including":[148],"ResNet-18":[150],"trained":[151],"on":[152],"CIFAR-10.":[153]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2023-10-26T00:00:00"}