{"id":"https://openalex.org/W3127730830","doi":null,"title":"Fantastic Four: Differentiable and Efficient Bounds on Singular Values of Convolution Layers","display_name":"Fantastic Four: Differentiable and Efficient Bounds on Singular Values of Convolution Layers","publication_year":2021,"publication_date":"2021-05-03","ids":{"openalex":"https://openalex.org/W3127730830","mag":"3127730830"},"language":"en","primary_location":{"id":"mag:3127730830","doi":null,"is_oa":false,"landing_page_url":"https://openreview.net/pdf?id=JCRblSgs34Z","pdf_url":null,"source":{"id":"https://openalex.org/S4306419637","display_name":"International Conference on Learning Representations","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"International Conference on Learning Representations","raw_type":null},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Sahil Singla","orcid":"https://orcid.org/0000-0002-8800-6479","observed_orcids":["https://orcid.org/0000-0002-8800-6479"]},"institutions":[{"id":"https://openalex.org/I66946132","display_name":"University of Maryland, College Park","ror":"https://ror.org/047s2c258","country_code":"US","type":"education","lineage":["https://openalex.org/I66946132"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sahil Singla","raw_affiliation_strings":["University of Maryland - College Park"],"raw_orcid":"https://orcid.org/0000-0002-8800-6479","affiliations":[{"raw_affiliation_string":"University of Maryland - College Park","institution_ids":["https://openalex.org/I66946132"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5025450606","display_name":"Soheil Feizi","orcid":"https://orcid.org/0000-0003-0944-8242","observed_orcids":["https://orcid.org/0000-0003-0944-8242","https://orcid.org/0000-0003-0493-4938"]},"institutions":[{"id":"https://openalex.org/I66946132","display_name":"University of Maryland, College Park","ror":"https://ror.org/047s2c258","country_code":"US","type":"education","lineage":["https://openalex.org/I66946132"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Soheil Feizi","raw_affiliation_strings":["University of Maryland - College Park"],"raw_orcid":"https://orcid.org/0000-0003-0944-8242","affiliations":[{"raw_affiliation_string":"University of Maryland - College Park","institution_ids":["https://openalex.org/I66946132"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I66946132"],"apc_list":null,"apc_paid":null,"fwci":0.3761,"has_fulltext":false,"cited_by_count":4,"citation_normalized_percentile":{"value":0.55473051,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":92,"max":95},"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":0.5375999808311462,"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":0.5375999808311462,"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":0.2524000108242035,"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/T10320","display_name":"Neural Networks and Applications","score":0.11879999935626984,"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/spectral-norm-bounds","display_name":"spectral norm bounds","score":0.8420000076293945},{"id":"https://openalex.org/keywords/convolutional-layers","display_name":"convolutional layers","score":0.8050000071525574},{"id":"https://openalex.org/keywords/lipschitz-constant","display_name":"Lipschitz constant","score":0.7620000243186951},{"id":"https://openalex.org/keywords/hessian-eigenvalues","display_name":"Hessian eigenvalues","score":0.9179999828338623},{"id":"https://openalex.org/keywords/generalization","display_name":"generalization","score":0.6589999794960022},{"id":"https://openalex.org/keywords/robustness","display_name":"robustness","score":0.9480000138282776}],"concepts":[{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.6265082955360413},{"id":"https://openalex.org/C202615002","wikidata":"https://www.wikidata.org/wiki/Q783507","display_name":"Differentiable function","level":2,"score":0.5947234034538269},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5698229670524597},{"id":"https://openalex.org/C190502265","wikidata":"https://www.wikidata.org/wiki/Q17069496","display_name":"MNIST database","level":3,"score":0.4695081412792206},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.4653182029724121},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.45774880051612854},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.44214576482772827},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.41113513708114624},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3383251428604126},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.324876606464386},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.20862606167793274},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.1236230731010437},{"id":"https://openalex.org/C55493867","wikidata":"https://www.wikidata.org/wiki/Q7094","display_name":"Biochemistry","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/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0},{"id":"https://openalex.org/C104317684","wikidata":"https://www.wikidata.org/wiki/Q7187","display_name":"Gene","level":2,"score":0.0},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"mag:3127730830","doi":null,"is_oa":false,"landing_page_url":"https://openreview.net/pdf?id=JCRblSgs34Z","pdf_url":null,"source":{"id":"https://openalex.org/S4306419637","display_name":"International Conference on Learning Representations","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":"International Conference on Learning Representations","raw_type":null}],"best_oa_location":null,"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":32,"referenced_works":["https://openalex.org/W194249466","https://openalex.org/W1525783482","https://openalex.org/W1981745143","https://openalex.org/W2108598243","https://openalex.org/W2112594540","https://openalex.org/W2112796928","https://openalex.org/W2132105090","https://openalex.org/W2144513243","https://openalex.org/W2163605009","https://openalex.org/W2168148636","https://openalex.org/W2194775991","https://openalex.org/W2618235498","https://openalex.org/W2624413595","https://openalex.org/W2709553318","https://openalex.org/W2753542656","https://openalex.org/W2785678896","https://openalex.org/W2796892552","https://openalex.org/W2799107510","https://openalex.org/W2803850896","https://openalex.org/W2803968356","https://openalex.org/W2808460953","https://openalex.org/W2913968085","https://openalex.org/W2945291039","https://openalex.org/W2947989684","https://openalex.org/W2962851953","https://openalex.org/W2962879692","https://openalex.org/W2963998105","https://openalex.org/W2964153729","https://openalex.org/W2964159205","https://openalex.org/W2964294232","https://openalex.org/W3035522104","https://openalex.org/W3118608800"],"related_works":["https://openalex.org/W3159623598","https://openalex.org/W4404673405","https://openalex.org/W4377081161","https://openalex.org/W4283724713","https://openalex.org/W4378501774","https://openalex.org/W3034411541","https://openalex.org/W2529714286","https://openalex.org/W4391505365","https://openalex.org/W2944820194","https://openalex.org/W4293243393"],"abstract_inverted_index":{"In":[0,71],"deep":[1,40,225],"neural":[2,203],"networks,":[3],"the":[4,8,14,18,47,76,89,120,144,156,171,192,200,213],"spectral":[5,48,90,172,179,217],"norm":[6,19,28,49,91,173],"of":[7,10,20,39,50,78,92,114,124,138,158,174,199,202,215,224],"Jacobian":[9],"a":[11,21,93,130,163],"layer":[12],"bounds":[13,87,100,116,141,161],"factor":[15,132],"by":[16,81,129],"which":[17],"signal":[22],"changes":[23],"during":[24,108],"forward/backward":[25],"propagation.":[26],"Spectral":[27],"regularizations":[29],"have":[30],"been":[31],"shown":[32],"to":[33,45,154,189],"improve":[34],"generalization,":[35],"robustness":[36,223],"and":[37,103,166,185,209,222],"optimization":[38],"learning":[41],"methods.":[42],"Existing":[43],"methods":[44],"compute":[46],"convolution":[51,97,149,175],"layers":[52],"either":[53,191],"rely":[54],"on":[55,88,134,148,170,207],"heuristics":[56],"that":[57],"are":[58,66,101],"efficient":[59,167],"in":[60,118,219],"computation":[61],"but":[62,68],"lack":[63],"guarantees":[64],"or":[65,195],"theoretically-sound":[67],"computationally":[69],"expensive.":[70],"this":[72],"work,":[73],"we":[74,152,211],"obtain":[75],"best":[77],"both":[79],"worlds":[80],"deriving":[82],"{\\it":[83],"four}":[84],"provable":[85],"upper":[86,168],"standard":[94],"2D":[95],"multi-channel":[96],"layer.":[98],"These":[99],"differentiable":[102,165],"can":[104,142,186],"be":[105,187],"computed":[106],"efficiently":[107],"training":[109],"with":[110],"negligible":[111],"overhead.":[112],"One":[113],"these":[115,139,159],"is":[117,181],"fact":[119],"popular":[121],"heuristic":[122],"method":[123],"Miyato":[125],"et":[126],"al.":[127],"(multiplied":[128],"constant":[131,194],"depending":[133,147],"filter":[135],"sizes).":[136],"Each":[137],"four":[140,160],"achieve":[143],"tightest":[145],"gap":[146],"filters.":[150],"Thus,":[151],"propose":[153],"use":[155],"minimum":[157],"as":[162],"tight,":[164],"bound":[169,180,190,218],"layers.":[176],"Moreover,":[177],"our":[178,216],"an":[182],"effective":[183],"regularizer":[184],"used":[188],"lipschitz":[193],"curvature":[196],"values":[197],"(eigenvalues":[198],"Hessian)":[201],"networks.":[204,226],"Through":[205],"experiments":[206],"MNIST":[208],"CIFAR-10,":[210],"demonstrate":[212],"effectiveness":[214],"improving":[220],"generalization":[221]},"counts_by_year":[{"year":2023,"cited_by_count":2},{"year":2021,"cited_by_count":2}],"updated_date":"2026-10-16T09:59:22.235173","created_date":"2021-02-15T00:00:00"}