{"id":"https://openalex.org/W3157298807","doi":"https://doi.org/10.1137/21m1415121","title":"Binary Classification of Gaussian Mixtures: Abundance of Support Vectors, Benign Overfitting, and Regularization","display_name":"Binary Classification of Gaussian Mixtures: Abundance of Support Vectors, Benign Overfitting, and Regularization","publication_year":2022,"publication_date":"2022-03-01","ids":{"openalex":"https://openalex.org/W3157298807","doi":"https://doi.org/10.1137/21m1415121","mag":"3157298807"},"language":"en","primary_location":{"id":"doi:10.1137/21m1415121","is_oa":true,"landing_page_url":"https://doi.org/10.1137/21m1415121","pdf_url":null,"source":{"id":"https://openalex.org/S4210229561","display_name":"SIAM Journal on Mathematics of Data Science","issn_l":"2577-0187","issn":["2577-0187"],"is_oa":true,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320508","host_organization_name":"Society for Industrial and Applied Mathematics","host_organization_lineage":["https://openalex.org/P4310320508"],"host_organization_lineage_names":["Society for Industrial and Applied Mathematics"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"SIAM Journal on Mathematics of Data Science","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1137/21m1415121","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100360159","display_name":"Ke Wang","orcid":"https://orcid.org/0000-0003-4501-5102"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ke Wang","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0003-4501-5102","affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5024812488","display_name":"Christos Thrampoulidis","orcid":"https://orcid.org/0000-0001-9053-9365"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Christos Thrampoulidis","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.6991,"has_fulltext":false,"cited_by_count":14,"citation_normalized_percentile":{"value":0.72957153,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":98},"biblio":{"volume":"4","issue":"1","first_page":"260","last_page":"284"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11901","display_name":"Bayesian Methods and Mixture Models","score":0.9969000220298767,"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/T11901","display_name":"Bayesian Methods and Mixture Models","score":0.9969000220298767,"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":0.9962999820709229,"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/T11447","display_name":"Blind Source Separation Techniques","score":0.9957000017166138,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/overfitting","display_name":"Overfitting","score":0.7703981399536133},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.5693908929824829},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.5656017065048218},{"id":"https://openalex.org/keywords/support-vector-machine","display_name":"Support vector machine","score":0.5579967498779297},{"id":"https://openalex.org/keywords/estimator","display_name":"Estimator","score":0.5233297348022461},{"id":"https://openalex.org/keywords/covariance","display_name":"Covariance","score":0.4838065505027771},{"id":"https://openalex.org/keywords/feature-vector","display_name":"Feature vector","score":0.47959935665130615},{"id":"https://openalex.org/keywords/binary-classification","display_name":"Binary classification","score":0.4724324643611908},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.44428738951683044},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4345533847808838},{"id":"https://openalex.org/keywords/binary-number","display_name":"Binary number","score":0.41703614592552185},{"id":"https://openalex.org/keywords/norm","display_name":"Norm (philosophy)","score":0.4159852862358093},{"id":"https://openalex.org/keywords/gaussian","display_name":"Gaussian","score":0.41168683767318726},{"id":"https://openalex.org/keywords/applied-mathematics","display_name":"Applied mathematics","score":0.3934780955314636},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3845794200897217},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.34787261486053467},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.3403344750404358},{"id":"https://openalex.org/keywords/statistics","display_name":"Statistics","score":0.2273753583431244},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.13137578964233398}],"concepts":[{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.7703981399536133},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.5693908929824829},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5656017065048218},{"id":"https://openalex.org/C12267149","wikidata":"https://www.wikidata.org/wiki/Q282453","display_name":"Support vector machine","level":2,"score":0.5579967498779297},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.5233297348022461},{"id":"https://openalex.org/C178650346","wikidata":"https://www.wikidata.org/wiki/Q201984","display_name":"Covariance","level":2,"score":0.4838065505027771},{"id":"https://openalex.org/C83665646","wikidata":"https://www.wikidata.org/wiki/Q42139305","display_name":"Feature vector","level":2,"score":0.47959935665130615},{"id":"https://openalex.org/C66905080","wikidata":"https://www.wikidata.org/wiki/Q17005494","display_name":"Binary classification","level":3,"score":0.4724324643611908},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.44428738951683044},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4345533847808838},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.41703614592552185},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.4159852862358093},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.41168683767318726},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3934780955314636},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3845794200897217},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.34787261486053467},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3403344750404358},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.2273753583431244},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.13137578964233398},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.0},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1137/21m1415121","is_oa":true,"landing_page_url":"https://doi.org/10.1137/21m1415121","pdf_url":null,"source":{"id":"https://openalex.org/S4210229561","display_name":"SIAM Journal on Mathematics of Data Science","issn_l":"2577-0187","issn":["2577-0187"],"is_oa":true,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320508","host_organization_name":"Society for Industrial and Applied Mathematics","host_organization_lineage":["https://openalex.org/P4310320508"],"host_organization_lineage_names":["Society for Industrial and Applied Mathematics"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"SIAM Journal on Mathematics of Data Science","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1137/21m1415121","is_oa":true,"landing_page_url":"https://doi.org/10.1137/21m1415121","pdf_url":null,"source":{"id":"https://openalex.org/S4210229561","display_name":"SIAM Journal on Mathematics of Data Science","issn_l":"2577-0187","issn":["2577-0187"],"is_oa":true,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320508","host_organization_name":"Society for Industrial and Applied Mathematics","host_organization_lineage":["https://openalex.org/P4310320508"],"host_organization_lineage_names":["Society for Industrial and Applied Mathematics"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"SIAM Journal on Mathematics of Data Science","raw_type":"journal-article"},"sustainable_development_goals":[{"display_name":"Reduced inequalities","score":0.7300000190734863,"id":"https://metadata.un.org/sdg/10"}],"awards":[{"id":"https://openalex.org/G6639552142","display_name":"CIF: Small: Fundamental Limits of Empirical Risk Minimization in High Dimensions: A Unifying Gaussian Processes Approach","funder_award_id":"2009030","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"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":59,"referenced_works":["https://openalex.org/W607505555","https://openalex.org/W835905657","https://openalex.org/W1537967768","https://openalex.org/W1554944419","https://openalex.org/W1689484996","https://openalex.org/W1982032418","https://openalex.org/W1993146053","https://openalex.org/W2029394273","https://openalex.org/W2032291279","https://openalex.org/W2040745222","https://openalex.org/W2068558570","https://openalex.org/W2161388792","https://openalex.org/W2163605009","https://openalex.org/W2168885649","https://openalex.org/W2550848904","https://openalex.org/W2604272474","https://openalex.org/W2752885492","https://openalex.org/W2787248994","https://openalex.org/W2787348029","https://openalex.org/W2907127169","https://openalex.org/W2911742574","https://openalex.org/W2923764619","https://openalex.org/W2948695458","https://openalex.org/W2950220847","https://openalex.org/W2963017107","https://openalex.org/W2963060833","https://openalex.org/W2963742538","https://openalex.org/W2964312599","https://openalex.org/W2965772785","https://openalex.org/W2967536008","https://openalex.org/W2970731147","https://openalex.org/W2971363918","https://openalex.org/W2989169642","https://openalex.org/W2994081359","https://openalex.org/W2996067004","https://openalex.org/W3003297907","https://openalex.org/W3013588718","https://openalex.org/W3014316192","https://openalex.org/W3018252856","https://openalex.org/W3021058579","https://openalex.org/W3025991882","https://openalex.org/W3034892535","https://openalex.org/W3035703482","https://openalex.org/W3046705928","https://openalex.org/W3090091013","https://openalex.org/W3096666815","https://openalex.org/W3111350549","https://openalex.org/W3111605110","https://openalex.org/W3112962553","https://openalex.org/W3127328563","https://openalex.org/W3138832176","https://openalex.org/W3145128584","https://openalex.org/W3151137625","https://openalex.org/W3157730063","https://openalex.org/W3157731489","https://openalex.org/W3162172987","https://openalex.org/W3168158992","https://openalex.org/W3172624703","https://openalex.org/W4285719527"],"related_works":["https://openalex.org/W1996541855","https://openalex.org/W2767651786","https://openalex.org/W3013515612","https://openalex.org/W2136184105","https://openalex.org/W2187500075","https://openalex.org/W2345184372","https://openalex.org/W2384093694","https://openalex.org/W2348964713","https://openalex.org/W2393055924","https://openalex.org/W2994154139"],"abstract_inverted_index":{"Deep":[0],"neural":[1,59],"networks":[2],"generalize":[3],"well":[4],"despite":[5],"being":[6,10],"exceedingly":[7],"overparameterized":[8],"and":[9,41,110,141,197,244,258],"trained":[11],"without":[12],"explicit":[13],"regularization.":[14],"This":[15],"curious":[16],"phenomenon":[17],"has":[18],"inspired":[19],"extensive":[20],"research":[21],"activity":[22],"in":[23,93,155],"establishing":[24],"its":[25,245],"statistical":[26],"principles:":[27],"Under":[28],"what":[29],"conditions":[30,192,263],"is":[31],"it":[32],"observed?":[33],"How":[34],"do":[35],"these":[36],"depend":[37],"on":[38,42,121,178,193,198],"the":[39,43,95,99,122,164,168,179,183,186,194,199,238,242,248,266],"data":[40,233,249],"training":[44],"algorithm?":[45],"When":[46],"does":[47],"regularization":[48],"benefit":[49],"generalization?":[50],"While":[51],"such":[52],"questions":[53],"remain":[54],"wide":[55],"open":[56],"for":[57,105],"deep":[58],"nets,":[60],"recent":[61,119],"works":[62],"have":[63],"attempted":[64],"gaining":[65],"insights":[66],"by":[67,82,118],"studying":[68],"simpler,":[69],"often":[70],"linear,":[71],"models.":[72],"Our":[73],"paper":[74],"contributes":[75],"to":[76,138,146,162,167,219,229],"this":[77],"growing":[78],"line":[79],"of":[80,125,182,241,255],"work":[81],"examining":[83],"binary":[84],"linear":[85],"classification":[86,180],"under":[87,204,264],"a":[88,106,220,253],"generative":[89],"Gaussian":[90],"mixture":[91],"model":[92,222],"which":[94,205,265],"feature":[96,111],"vectors":[97],"take":[98],"form":[100],"${{\\it":[101,113],"x}}=\\pm{{\\eta}}+{{\\it":[102],"q}}$,":[103],"where":[104],"mean":[107],"vector":[108,133],"$\\eta$":[109],"noise":[112,226],"q}}":[114],"\\sim":[115],"\\mathcal{N}(0,{{\\Sigma}})$.":[116],"Motivated":[117],"results":[120,217,236],"implicit":[123],"bias":[124],"gradient":[126],"descent,":[127],"we":[128,150,173,188,261],"study":[129],"both":[130],"max-margin":[131],"support":[132],"machine":[134],"(SVM)":[135],"classifiers":[136,144],"(corresponding":[137,145],"logistic":[139],"loss)":[140],"min-norm":[142,169],"interpolating":[143,170,206,267],"least-squares":[147],"loss).":[148],"First,":[149],"leverage":[151],"an":[152],"idea":[153],"introduced":[154],"[V.":[156],"Muthukumar":[157],"et":[158],"al.,":[159],"arXiv:2005.08054,":[160],"2020a]":[161],"relate":[163],"SVM":[165],"solution":[166],"solution.":[171],"Second,":[172],"derive":[174],"novel":[175,190],"nonasymptotic":[176],"bounds":[177],"error":[181],"latter.":[184],"Combining":[185],"two,":[187],"present":[189],"sufficient":[191],"covariance":[195],"spectrum":[196],"signal-to-noise":[200],"ratio":[201],"(SNR)":[202],"$SNR={||{{\\eta}}||_2^4}/{{\\eta}}^T{{\\Sigma\\eta}}$":[203],"estimators":[207],"achieve":[208],"asymptotically":[209],"optimal":[210],"performance":[211],"as":[212],"overparameterization":[213],"increases.":[214],"Interestingly,":[215],"our":[216,235],"extend":[218],"noisy":[221],"with":[223,247],"constant":[224],"probability":[225],"flips.":[227],"Contrary":[228],"previously":[230],"studied":[231],"discriminative":[232],"models,":[234],"emphasize":[237],"crucial":[239],"role":[240],"SNR":[243],"interplay":[246],"covariance.":[250],"Finally,":[251],"via":[252],"combination":[254],"analytical":[256],"arguments":[257],"numerical":[259],"demonstrations":[260],"identify":[262],"estimator":[268],"performs":[269],"better":[270],"than":[271],"corresponding":[272],"regularized":[273],"estimates.":[274]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2024,"cited_by_count":3},{"year":2023,"cited_by_count":2},{"year":2022,"cited_by_count":1},{"year":2021,"cited_by_count":6},{"year":2019,"cited_by_count":1}],"updated_date":"2026-08-06T08:24:18.245995","created_date":"2025-10-10T00:00:00"}