{"id":"https://openalex.org/W4415482185","doi":"https://doi.org/10.1109/tsp.2025.3624791","title":"Regularized Top-$ k $: A Bayesian Framework for Gradient Sparsification","display_name":"Regularized Top-$ k $: A Bayesian Framework for Gradient Sparsification","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4415482185","doi":"https://doi.org/10.1109/tsp.2025.3624791"},"language":null,"primary_location":{"id":"doi:10.1109/tsp.2025.3624791","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tsp.2025.3624791","pdf_url":null,"source":{"id":"https://openalex.org/S168680287","display_name":"IEEE Transactions on Signal Processing","issn_l":"1053-587X","issn":["1053-587X","1941-0476"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Signal Processing","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5061064331","display_name":"Ali Bereyhi","orcid":"https://orcid.org/0000-0001-9565-6405"},"institutions":[{"id":"https://openalex.org/I185261750","display_name":"University of Toronto","ror":"https://ror.org/03dbr7087","country_code":"CA","type":"education","lineage":["https://openalex.org/I185261750"]}],"countries":["CA"],"is_corresponding":true,"raw_author_name":"Ali Bereyhi","raw_affiliation_strings":["Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 2E4, Canada","Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada"],"affiliations":[{"raw_affiliation_string":"Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 2E4, Canada","institution_ids":["https://openalex.org/I185261750"]},{"raw_affiliation_string":"Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada","institution_ids":["https://openalex.org/I185261750"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041434153","display_name":"Ben Liang","orcid":"https://orcid.org/0000-0002-1800-1322"},"institutions":[{"id":"https://openalex.org/I185261750","display_name":"University of Toronto","ror":"https://ror.org/03dbr7087","country_code":"CA","type":"education","lineage":["https://openalex.org/I185261750"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Ben Liang","raw_affiliation_strings":["Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 2E4, Canada","Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada"],"affiliations":[{"raw_affiliation_string":"Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 2E4, Canada","institution_ids":["https://openalex.org/I185261750"]},{"raw_affiliation_string":"Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada","institution_ids":["https://openalex.org/I185261750"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037720618","display_name":"Gary Boudreau","orcid":"https://orcid.org/0000-0003-3539-9624"},"institutions":[{"id":"https://openalex.org/I4210094041","display_name":"Ericsson (Canada)","ror":"https://ror.org/00nas2c56","country_code":"CA","type":"company","lineage":["https://openalex.org/I1306339040","https://openalex.org/I4210094041"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Gary Boudreau","raw_affiliation_strings":["Ericsson Canada, Ottawa, ON K2K 2V6, Canada","Ericsson Canada, Ottawa, Canada"],"affiliations":[{"raw_affiliation_string":"Ericsson Canada, Ottawa, ON K2K 2V6, Canada","institution_ids":["https://openalex.org/I4210094041"]},{"raw_affiliation_string":"Ericsson Canada, Ottawa, Canada","institution_ids":["https://openalex.org/I4210094041"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5071889415","display_name":"Ali Afana","orcid":"https://orcid.org/0000-0001-5253-2429"},"institutions":[{"id":"https://openalex.org/I4210094041","display_name":"Ericsson (Canada)","ror":"https://ror.org/00nas2c56","country_code":"CA","type":"company","lineage":["https://openalex.org/I1306339040","https://openalex.org/I4210094041"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Ali Afana","raw_affiliation_strings":["Ericsson Canada, Ottawa, ON K2K 2V6, Canada","Ericsson Canada, Ottawa, Canada"],"affiliations":[{"raw_affiliation_string":"Ericsson Canada, Ottawa, ON K2K 2V6, Canada","institution_ids":["https://openalex.org/I4210094041"]},{"raw_affiliation_string":"Ericsson Canada, Ottawa, Canada","institution_ids":["https://openalex.org/I4210094041"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5061064331"],"corresponding_institution_ids":["https://openalex.org/I185261750"],"apc_list":null,"apc_paid":null,"fwci":1.428,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.86724629,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":"73","issue":null,"first_page":"4463","last_page":"4478"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10052","display_name":"Medical Image Segmentation Techniques","score":0.9955000281333923,"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/T10052","display_name":"Medical Image Segmentation Techniques","score":0.9955000281333923,"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/T10719","display_name":"3D Shape Modeling and Analysis","score":0.9947999715805054,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10191","display_name":"Robotics and Sensor-Based Localization","score":0.993399977684021,"subfield":{"id":"https://openalex.org/subfields/2202","display_name":"Aerospace Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.6195999979972839},{"id":"https://openalex.org/keywords/posterior-probability","display_name":"Posterior probability","score":0.5284000039100647},{"id":"https://openalex.org/keywords/convergence","display_name":"Convergence (economics)","score":0.4918000102043152},{"id":"https://openalex.org/keywords/rate-of-convergence","display_name":"Rate of convergence","score":0.48100000619888306},{"id":"https://openalex.org/keywords/bayesian-probability","display_name":"Bayesian probability","score":0.45649999380111694},{"id":"https://openalex.org/keywords/prior-probability","display_name":"Prior probability","score":0.4424999952316284},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.4311999976634979},{"id":"https://openalex.org/keywords/inverse","display_name":"Inverse","score":0.38190001249313354},{"id":"https://openalex.org/keywords/inverse-problem","display_name":"Inverse problem","score":0.3806999921798706}],"concepts":[{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.6195999979972839},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.6000000238418579},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5514000058174133},{"id":"https://openalex.org/C57830394","wikidata":"https://www.wikidata.org/wiki/Q278079","display_name":"Posterior probability","level":3,"score":0.5284000039100647},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.4918000102043152},{"id":"https://openalex.org/C57869625","wikidata":"https://www.wikidata.org/wiki/Q1783502","display_name":"Rate of convergence","level":3,"score":0.48100000619888306},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.45649999380111694},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.4424999952316284},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.4311999976634979},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.4077000021934509},{"id":"https://openalex.org/C207467116","wikidata":"https://www.wikidata.org/wiki/Q4385666","display_name":"Inverse","level":2,"score":0.38190001249313354},{"id":"https://openalex.org/C135252773","wikidata":"https://www.wikidata.org/wiki/Q1567213","display_name":"Inverse problem","level":2,"score":0.3806999921798706},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.37959998846054077},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.3637999892234802},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.36309999227523804},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.36160001158714294},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3391999900341034},{"id":"https://openalex.org/C115680565","wikidata":"https://www.wikidata.org/wiki/Q5977448","display_name":"Gradient method","level":2,"score":0.30309998989105225},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.3012999892234802},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.2883000075817108},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2797999978065491},{"id":"https://openalex.org/C159694833","wikidata":"https://www.wikidata.org/wiki/Q2321565","display_name":"Iterative method","level":2,"score":0.27149999141693115},{"id":"https://openalex.org/C94835093","wikidata":"https://www.wikidata.org/wiki/Q3113333","display_name":"Data compression ratio","level":5,"score":0.27129998803138733},{"id":"https://openalex.org/C10494615","wikidata":"https://www.wikidata.org/wiki/Q17086765","display_name":"Proximal Gradient Methods","level":4,"score":0.2689000070095062},{"id":"https://openalex.org/C2779582901","wikidata":"https://www.wikidata.org/wiki/Q21013010","display_name":"Distributed learning","level":2,"score":0.2644999921321869},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.25209999084472656},{"id":"https://openalex.org/C170122806","wikidata":"https://www.wikidata.org/wiki/Q1914828","display_name":"Linear scale","level":2,"score":0.250900000333786}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tsp.2025.3624791","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tsp.2025.3624791","pdf_url":null,"source":{"id":"https://openalex.org/S168680287","display_name":"IEEE Transactions on Signal Processing","issn_l":"1053-587X","issn":["1053-587X","1941-0476"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Signal Processing","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":22,"referenced_works":["https://openalex.org/W2108598243","https://openalex.org/W2154518298","https://openalex.org/W2194775991","https://openalex.org/W2405578611","https://openalex.org/W2407022425","https://openalex.org/W2499292135","https://openalex.org/W2606891064","https://openalex.org/W2775776326","https://openalex.org/W2883780447","https://openalex.org/W2963125010","https://openalex.org/W2963163009","https://openalex.org/W2996979369","https://openalex.org/W3012125688","https://openalex.org/W3021654819","https://openalex.org/W3129329365","https://openalex.org/W4210869902","https://openalex.org/W4214711850","https://openalex.org/W4220826035","https://openalex.org/W4230874317","https://openalex.org/W4403210662","https://openalex.org/W4404037626","https://openalex.org/W4415482185"],"related_works":[],"abstract_inverted_index":{"Error":[0],"accumulation":[1,25],"is":[2,87,101,186],"effective":[3],"for":[4,35],"gradient":[5,11,50,85,166],"sparsification":[6,64,86,99,120,261],"in":[7,48,56,228],"distributed":[8,49,182,229],"settings:":[9],"initially-unselected":[10],"entries":[12,167],"are":[13],"eventually":[14],"selected":[15,37],"as":[16,28,89,103,126,235,237,253],"their":[17],"accumulated":[18,165],"error":[19,73],"exceeds":[20],"a":[21,29,62,104,118,127,195],"certain":[22],"level.":[23],"The":[24,75],"essentially":[26],"behaves":[27],"scaling":[30,71],"of":[31,45,72,77,130,156,220,231,239],"the":[32,36,43,68,96,108,157,163,199,207,218,245,254],"learning":[33,69],"rate":[34,70],"entries.":[38],"Although":[39],"this":[40,78,136,221],"property":[41],"prevents":[42],"slow-down":[44],"lateral":[46],"movements":[47],"descent,":[51],"it":[52,185],"can":[53,123],"deteriorate":[54],"convergence":[55],"some":[57],"settings.":[58],"This":[59],"work":[60],"proposes":[61],"novel":[63],"scheme":[65,79],"that":[66,188,252],"controls":[67],"accumulation.":[74],"development":[76],"follows":[80],"two":[81],"major":[82],"steps:":[83],"first,":[84],"formulated":[88],"an":[90],"inverse":[91],"probability":[92],"(inference)":[93],"problem,":[94],"and":[95],"Bayesian":[97],"optimal":[98],"mask":[100],"derived":[102],"maximum-a-posteriori":[105],"estimator.":[106],"Using":[107],"prior":[109],"distribution":[110],"inherited":[111],"from":[112,198],"TOP-<italic":[113,131,141,190,264],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[114,132,139,142,145,191,203,226,259,265],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">k</i>,":[115],"we":[116],"derive":[117],"new":[119],"algorithm":[121,137],"which":[122],"be":[124],"interpreted":[125],"regularized":[128],"form":[129],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">k</i>.":[133,266],"We":[134,173,215],"call":[135],"<italic":[138],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">regularized</i>":[140],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">k</i>":[143,192,204,227,260],"(REGTOP-<italic":[144],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">k</i>).":[146],"It":[147,160],"utilizes":[148],"past":[149],"aggregated":[150],"gradients":[151],"to":[152,206],"evaluate":[153],"posterior":[154,171],"statistics":[155],"next":[158],"aggregation.":[159],"then":[161],"prioritizes":[162],"local":[164],"based":[168],"on":[169,233,244],"these":[170],"statistics.":[172],"validate":[174],"our":[175],"derivation":[176],"through":[177],"various":[178],"numerical":[179,249],"experiments.":[180],"In":[181],"linear":[183],"regression,":[184],"observed":[187],"while":[189],"remains":[193],"at":[194,210],"fixed":[196],"distance":[197],"global":[200,208],"optimum,":[201],"REGTOP-<italic":[202,225,258],"converges":[205],"optimum":[209],"significantly":[211],"higher":[212],"compression":[213,255],"ratios.":[214],"further":[216],"demonstrate":[217],"generalization":[219],"observation":[222],"by":[223],"employing":[224],"training":[230],"ResNet-18":[232],"CIFAR-10,":[234],"well":[236],"fine-tuning":[238],"multiple":[240],"computer":[241],"vision":[242],"models":[243],"ImageNette":[246],"dataset.":[247],"Our":[248],"results":[250],"confirm":[251],"ratio":[256],"increases,":[257],"noticeably":[262],"outperforms":[263]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2025-10-23T00:00:00"}