A Method to Solve overflow or underflow errors Resulting from Activation functions in Convolutional Neural Networks

In this paper, a method is proposed to correct overflow errors of resulting values that can occur when the cross-entropy loss function is used as a softmax and ReLU as an activation function of the convolutional neural network. Since conventional softmax and cross-entropy functions include exponenti...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:International Information Institute (Tokyo). Information Ročník 20; číslo 11; s. 8151 - 8158
Hlavní autoři: Jeong, JunMo, Choi, SeJin, Kim, ChiYong
Médium: Journal Article
Jazyk:angličtina
Vydáno: Koganei International Information Institute 01.11.2017
Témata:
ISSN:1343-4500, 1344-8994
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract In this paper, a method is proposed to correct overflow errors of resulting values that can occur when the cross-entropy loss function is used as a softmax and ReLU as an activation function of the convolutional neural network. Since conventional softmax and cross-entropy functions include exponential and log computations, they have a problem of overflow or underflow errors, which produce a range of numbers such as convergence to zero or infinite divergence when a value of the final output layer becomes too large. In the present paper, the above problem is solved by transforming an equation without affecting the functions of the conventional softmax and cross-entropy functions. Thus, training can be achieved to correcting the overflow or underflow error problem regardless of a size of values in the final output value.
AbstractList In this paper, a method is proposed to correct overflow errors of resulting values that can occur when the cross-entropy loss function is used as a softmax and ReLU as an activation function of the convolutional neural network. Since conventional softmax and cross-entropy functions include exponential and log computations, they have a problem of overflow or underflow errors, which produce a range of numbers such as convergence to zero or infinite divergence when a value of the final output layer becomes too large. In the present paper, the above problem is solved by transforming an equation without affecting the functions of the conventional softmax and cross-entropy functions. Thus, training can be achieved to correcting the overflow or underflow error problem regardless of a size of values in the final output value.
Author Jeong, JunMo
Choi, SeJin
Kim, ChiYong
Author_xml – sequence: 1
  givenname: JunMo
  surname: Jeong
  fullname: Jeong, JunMo
– sequence: 2
  givenname: SeJin
  surname: Choi
  fullname: Choi, SeJin
– sequence: 3
  givenname: ChiYong
  surname: Kim
  fullname: Kim, ChiYong
BookMark eNotjttKAzEURYNUsNb-Q8DngTNJppM8luINqoKX55KZnNGpY47mMv1929qntTdsFvuSTTx5PGPTUipVaGPU5JhloSqACzaPcQsAEsAsoJyyuOSPmD7J8UT8lYYROY0YuoF2nALP3p0KhkAh8heMeUi9_-BdoG--bFM_2tST51327SFE3nu-Ij_SkA_dDvwJczgi7Sh8xSt23tkh4vzEGXu_vXlb3Rfr57uH1XJd-FJUqTB1J4VEY1UDQkFVCtWoTpTCmbKprYRGK7RYt9qBcChquUAFC9RtbZ3WTs7Y9b_3J9Bvxpg2W8phfyhuBOw9-7E28g-RLlwE
ContentType Journal Article
Copyright Copyright International Information Institute Nov 2017
Copyright_xml – notice: Copyright International Information Institute Nov 2017
DBID 7SC
8FD
8FE
8FG
ABUWG
AFKRA
ARAPS
BENPR
BGLVJ
BVBZV
CCPQU
DWQXO
HCIFZ
JQ2
L7M
L~C
L~D
P5Z
P62
PHGZM
PHGZT
PKEHL
PQEST
PQGLB
PQQKQ
PQUKI
PRINS
DatabaseName Computer and Information Systems Abstracts
Technology Research Database
ProQuest SciTech Collection
ProQuest Technology Collection
ProQuest Central (Alumni)
ProQuest Central UK/Ireland
Advanced Technologies & Computer Science Collection
ProQuest Central
Technology collection
East & South Asia Database
ProQuest One Community College
ProQuest Central
SciTech Premium Collection
ProQuest Computer Science Collection
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
Advanced Technologies & Aerospace Database
ProQuest Advanced Technologies & Aerospace Collection
ProQuest Central Premium
ProQuest One Academic
ProQuest One Academic Middle East (New)
ProQuest One Academic Eastern Edition (DO NOT USE)
ProQuest One Applied & Life Sciences
ProQuest One Academic (retired)
ProQuest One Academic UKI Edition
ProQuest Central China
DatabaseTitle Technology Collection
Technology Research Database
Computer and Information Systems Abstracts – Academic
ProQuest One Academic Middle East (New)
ProQuest Advanced Technologies & Aerospace Collection
ProQuest Computer Science Collection
Computer and Information Systems Abstracts
ProQuest Central (Alumni Edition)
SciTech Premium Collection
ProQuest One Community College
ProQuest Central China
ProQuest Central
ProQuest One Applied & Life Sciences
ProQuest Central Korea
ProQuest Central (New)
Advanced Technologies Database with Aerospace
Advanced Technologies & Aerospace Collection
ProQuest One Academic Eastern Edition
ProQuest Technology Collection
ProQuest SciTech Collection
East & South Asia Database
Computer and Information Systems Abstracts Professional
Advanced Technologies & Aerospace Database
ProQuest One Academic UKI Edition
ProQuest One Academic
ProQuest One Academic (New)
DatabaseTitleList Technology Collection
Database_xml – sequence: 1
  dbid: BENPR
  name: ProQuest Central
  url: https://www.proquest.com/central
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
EISSN 1344-8994
EndPage 8158
GroupedDBID 7SC
8FD
8FE
8FG
ABUWG
AFKRA
ALMA_UNASSIGNED_HOLDINGS
ARAPS
BENPR
BGLVJ
BPHCQ
BVBZV
CCPQU
DWQXO
HCIFZ
JQ2
L7M
L~C
L~D
P62
PHGZM
PHGZT
PKEHL
PQEST
PQGLB
PQQKQ
PQUKI
PRINS
PROAC
ID FETCH-LOGICAL-n125t-97f323e9a4b02405124b4f212d91b7a30b84eae7c8d02de2736e406e8c7ad88d3
IEDL.DBID P5Z
ISSN 1343-4500
IngestDate Sun Jul 13 03:47:17 EDT 2025
IsPeerReviewed false
IsScholarly false
Issue 11
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-n125t-97f323e9a4b02405124b4f212d91b7a30b84eae7c8d02de2736e406e8c7ad88d3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
PQID 2021240689
PQPubID 936334
PageCount 8
ParticipantIDs proquest_journals_2021240689
PublicationCentury 2000
PublicationDate 20171101
PublicationDateYYYYMMDD 2017-11-01
PublicationDate_xml – month: 11
  year: 2017
  text: 20171101
  day: 01
PublicationDecade 2010
PublicationPlace Koganei
PublicationPlace_xml – name: Koganei
PublicationTitle International Information Institute (Tokyo). Information
PublicationYear 2017
Publisher International Information Institute
Publisher_xml – name: International Information Institute
SSID ssj0003009601
Score 1.6699216
Snippet In this paper, a method is proposed to correct overflow errors of resulting values that can occur when the cross-entropy loss function is used as a softmax and...
SourceID proquest
SourceType Aggregation Database
StartPage 8151
SubjectTerms Activation
Algorithms
Artificial neural networks
Classification
Divergence
Entropy
Error correction
Methods
Neural networks
Neurons
Title A Method to Solve overflow or underflow errors Resulting from Activation functions in Convolutional Neural Networks
URI https://www.proquest.com/docview/2021240689
Volume 20
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVPQU
  databaseName: Advanced Technologies & Aerospace Database
  customDbUrl:
  eissn: 1344-8994
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0003009601
  issn: 1343-4500
  databaseCode: P5Z
  dateStart: 20111101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/hightechjournals
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: East & South Asia Database
  customDbUrl:
  eissn: 1344-8994
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0003009601
  issn: 1343-4500
  databaseCode: BVBZV
  dateStart: 20111101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/eastsouthasia
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: ProQuest Central
  customDbUrl:
  eissn: 1344-8994
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0003009601
  issn: 1343-4500
  databaseCode: BENPR
  dateStart: 20111101
  isFulltext: true
  titleUrlDefault: https://www.proquest.com/central
  providerName: ProQuest
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV07T8MwELagZWDhIUA8SuWBNVISO008oVC1YqGqCkgVSxXbF1SpiiEO5e_jc1MYkFiYoihD4td9913u7iPkplCJHshYBgAMCQobBEXpzlXKhHT8WRSRUF5sIp1MsvlcTNuAm23TKrc20RtqbRTGyJGkR4g-mbh9ew9QNQr_rrYSGruki10SULphmrx8x1iYd9A3nIuzgCdh-MvkehwZH_73C47IQetB0nyz5MdkB6oTYnP64LWgaWPoo1mtgWJuZrkyn9TUFCvFNjdQ16a2dAYWMwmrV4r1JTRXW5UzikDn9yJdVnRoqnW7N90rsZGHv_jMcXtKnsejp-F90OopBJVzY5pApCWLGYiCS-xs5qCeS166QWkRybRgocw4FJCqTIexBufYDMCNFzKVFjrLNDsjncpUcE6olFLwOEwgDksOLJHCES9ZRqVOojSU6oL0tnO4aA-FXfxM4OXfj6_Ifozo6Uv-eqTT1B9wTfbUulnauk-6d6PJdNb3a_0FzQW3Gw
linkProvider ProQuest
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMw1V1LS8NAEB5qK-jFByo-qu5BvQWSbF578NBWS0sfiNZSvNRsMpFCSTSJLf4pf6O721QPgrcePIUQ2GWZnfnm28zsB3DhB3bocJNriFQSFOpofiT8yqWMC_7MfIMFSmzC7fe90YjdleBz2QsjyyqXMVEF6jAJ5Bm5JOmGRB-PFRWUHfyYC36WXbdvhDEvTbN5O2i0tEJCQIsFcucacyNqUmS-xeVlXgLdLG5FYqSQGdz1qc49C310Ay_UzRAFljsoJkEvcP3Q80Iqxr16fdOkSpX8m1tIdqxBxXNEblKGSn1Yfxp-n-pQRQkWLM-immXr-q8gr5Cruf3f1rwDW0WOTGqLTb0LJYz3IKuRnlK7JnlCHpLpDImsPo2myZwkKZG9cIsXTNMkzcg9ZrJWMn4hsoOG1IKljhuRUK68jUxi0kjiWeF9Ykp5VYl6qNr4bB8eV7L0AyjHSYyHQDjnzDJ1G009spDanAlqySMjCm3D1XlwBNWlzcaF22fjH4Md__35HDZag1533G33OyewacpcQTU4VqGcp-94CuvBLJ9k6Vmxwwg8r9rmXyLvEwE
linkToPdf http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMw1V1NS8MwGA5zinjxAxU_puagx2LbtGtzEBmbwzEdww8YXmaTvJHBaLWtG_41f51J2upB8LaDp1IKTcP78eRJ3zcPQqcR90WTucwCIJqgkKYVSRVXAaFM8WcaOZQbsYlgMAhHIzqsoc-qF0aXVVY50SRqkXC9R65JuqPRJ6TnsiyLGHa6l69vllaQ0n9aKzmNwkX68DFX9C276HWUrc9ct3v10L62SoUBK1bAnls0kMQlQCOP6bO-FPh5zJNqIEEdFkTEZqEHEQQ8FLYrQEF9E9Q3QMiDSIShIOq9S2g5UBxTlxMO_afv_R1iyEHB9zxieb5t_0r3BsO6G_959ptovVw541bh6luoBvE2ylr41mhg4zzB98l0BljXpMppMsdJinWHXHEDaZqkGb6DTFdQxi9Y99XgFq_U3bAGeBODeBLjdhLPyphUQ-oDTMzFVMxnO-hxIfPcRfU4iWEPYcYY9VzbB9eWHhCfUUU4mXSk8J3AZnwfNSr7jctkkI1_jHfw9-MTtKqsOr7pDfqHaM3VCwjT9dhA9Tx9hyO0wmf5JEuPjath9Lxo634BarQaZw
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=A+Method+to+Solve+overflow+or+underflow+errors+Resulting+from+Activation+functions+in+Convolutional+Neural+Networks&rft.jtitle=International+Information+Institute+%28Tokyo%29.+Information&rft.au=Jeong%2C+JunMo&rft.au=Choi%2C+SeJin&rft.au=Kim%2C+ChiYong&rft.date=2017-11-01&rft.pub=International+Information+Institute&rft.issn=1343-4500&rft.eissn=1344-8994&rft.volume=20&rft.issue=11&rft.spage=8151&rft.epage=8158&rft.externalDBID=HAS_PDF_LINK
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1343-4500&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1343-4500&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1343-4500&client=summon