{"id":"https://openalex.org/W7133324107","doi":"https://doi.org/10.48550/arxiv.2603.00951","title":"When Does Margin Clamping Affect Training Variance? Dataset-Dependent Effects in Contrastive Forward-Forward Learning","display_name":"When Does Margin Clamping Affect Training Variance? Dataset-Dependent Effects in Contrastive Forward-Forward Learning","publication_year":2026,"publication_date":"2026-03-01","ids":{"openalex":"https://openalex.org/W7133324107","doi":"https://doi.org/10.48550/arxiv.2603.00951"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.00951","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.00951","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.00951","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127965809","display_name":"Joshua Steier","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Steier, Joshua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5127965809"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.7407000064849854,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.7407000064849854,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.03610000014305115,"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":0.0357000008225441,"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/clamping","display_name":"Clamping","score":0.515999972820282},{"id":"https://openalex.org/keywords/variance","display_name":"Variance (accounting)","score":0.47029998898506165},{"id":"https://openalex.org/keywords/margin","display_name":"Margin (machine learning)","score":0.4250999987125397},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.42309999465942383},{"id":"https://openalex.org/keywords/contrast","display_name":"Contrast (vision)","score":0.3950999975204468},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.39489999413490295},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.3278000056743622},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.31380000710487366}],"concepts":[{"id":"https://openalex.org/C84111939","wikidata":"https://www.wikidata.org/wiki/Q5125465","display_name":"Clamping","level":2,"score":0.515999972820282},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4772999882698059},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.47029998898506165},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4648999869823456},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.44850000739097595},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.4250999987125397},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.42309999465942383},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.39649999141693115},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.3950999975204468},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.39489999413490295},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.3278000056743622},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3246999979019165},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.31380000710487366},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.3107999861240387},{"id":"https://openalex.org/C106195933","wikidata":"https://www.wikidata.org/wiki/Q7847935","display_name":"Truncation (statistics)","level":2,"score":0.30480000376701355},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3005000054836273},{"id":"https://openalex.org/C68060419","wikidata":"https://www.wikidata.org/wiki/Q40754","display_name":"Subtraction","level":2,"score":0.3003999888896942},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.29600000381469727},{"id":"https://openalex.org/C207821765","wikidata":"https://www.wikidata.org/wiki/Q405372","display_name":"Instability","level":2,"score":0.2903999984264374},{"id":"https://openalex.org/C2776035688","wikidata":"https://www.wikidata.org/wiki/Q1606558","display_name":"Affect (linguistics)","level":2,"score":0.28949999809265137},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.28839999437332153},{"id":"https://openalex.org/C2776161997","wikidata":"https://www.wikidata.org/wiki/Q846600","display_name":"Clamp","level":3,"score":0.2831999957561493},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.2727000117301941},{"id":"https://openalex.org/C129844170","wikidata":"https://www.wikidata.org/wiki/Q41299","display_name":"Quadratic equation","level":2,"score":0.27059999108314514},{"id":"https://openalex.org/C122048520","wikidata":"https://www.wikidata.org/wiki/Q2913954","display_name":"Percentile","level":2,"score":0.26429998874664307}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.00951","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.00951","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.00951","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.00951","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/8","score":0.4449465870857239,"display_name":"Decent work and economic growth"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Contrastive":[0],"Forward-Forward":[1],"(CFF)":[2],"learning":[3],"trains":[4],"Vision":[5],"Transformers":[6],"layer":[7,9],"by":[8],"against":[10],"supervised":[11],"contrastive":[12,42],"objectives.":[13],"CFF":[14],"training":[15],"can":[16],"be":[17],"sensitive":[18],"to":[19,109,122,183,198,210],"random":[20],"seed,":[21],"but":[22],"the":[23,37,41,61,64,69,140,167,174,199,214,226],"sources":[24],"of":[25,97],"this":[26,204],"instability":[27],"are":[28],"poorly":[29],"understood.":[30],"We":[31,54],"focus":[32],"on":[33,125,169],"one":[34],"implementation":[35],"detail:":[36],"positive-pair":[38,143],"margin":[39,62],"in":[40,93],"loss":[43],"is":[44,66,160],"applied":[45],"through":[46],"saturating":[47],"similarity":[48],"clamping,":[49],"$\\min(s":[50],"+":[51],"m,\\,":[52],"1)$.":[53],"prove":[55],"that":[56],"an":[57],"alternative":[58],"formulation,":[59],"subtracting":[60],"after":[63],"log-probability,":[65],"gradient-neutral":[67,200],"under":[68,185],"mean-over-positives":[70],"reduction.":[71],"On":[72],"CIFAR-10":[73],"($2":[74],"\\times":[75],"2$":[76],"factorial,":[77],"$n{=}7$":[78],"seeds":[79],"per":[80,145,195],"cell),":[81],"clamping":[82,130],"produces":[83,131],"$5.90\\times$":[84],"higher":[85],"pooled":[86],"test-accuracy":[87],"variance":[88,175,205],"($p{=}0.003$)":[89],"with":[90,173,191],"no":[91,208],"difference":[92],"mean":[94,211],"accuracy.":[95,212],"Analyses":[96],"clamp":[98,216],"activation":[99,217],"rates,":[100],"layerwise":[101],"gradient":[102,111],"norms,":[103],"and":[104,128],"a":[105,170,221],"reduced-margin":[106],"probe":[107],"point":[108],"saturation-driven":[110],"truncation":[112],"at":[113,180,207],"early":[114],"layers.":[115],"The":[116],"effect":[117],"does":[118],"not":[119],"transfer":[120],"cleanly":[121],"other":[123],"datasets:":[124],"CIFAR-100,":[126],"SVHN,":[127],"Fashion-MNIST,":[129],"equal":[132],"or":[133],"lower":[134],"variance.":[135],"Two":[136],"factors":[137],"account":[138],"for":[139,224],"discrepancy.":[141],"First,":[142],"density":[144],"batch":[146],"controls":[147],"how":[148],"often":[149],"saturation":[150],"occurs.":[151],"Second,":[152],"task":[153],"difficulty":[154,164],"compresses":[155],"seed-to-seed":[156],"spread":[157],"when":[158],"accuracy":[159,182],"high.":[161],"An":[162],"SVHN":[163],"sweep":[165],"confirms":[166],"interaction":[168],"single":[171],"dataset,":[172],"ratio":[176],"moving":[177],"from":[178],"$0.25\\times$":[179],"high":[181],"$16.73\\times$":[184],"aggressive":[186],"augmentation.":[187],"In":[188],"moderate-accuracy":[189],"regimes":[190],"many":[192],"same-class":[193],"pairs":[194],"batch,":[196],"switching":[197],"subtraction":[201],"reference":[202],"removes":[203],"inflation":[206],"cost":[209],"Measuring":[213],"layer-0":[215],"rate":[218],"serves":[219],"as":[220],"simple":[222],"check":[223],"whether":[225],"problem":[227],"applies.":[228]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-04T00:00:00"}