Continuously Constructive Deep Neural Networks

Traditionally, deep learning algorithms update the network weights, whereas the network architecture is chosen manually using a process of trial and error. In this paper, we propose two novel approaches that automatically update the network structure while also learning its weights. The novelty of o...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transaction on neural networks and learning systems Ročník 31; číslo 4; s. 1124 - 1133
Hlavní autoři: Irsoy, Ozan, Alpaydin, Ethem
Médium: Journal Article
Jazyk:angličtina
Vydáno: United States IEEE 01.04.2020
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:
ISSN:2162-237X, 2162-2388, 2162-2388
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract Traditionally, deep learning algorithms update the network weights, whereas the network architecture is chosen manually using a process of trial and error. In this paper, we propose two novel approaches that automatically update the network structure while also learning its weights. The novelty of our approach lies in our parameterization, where the depth, or additional complexity, is encapsulated continuously in the parameter space through control parameters that add additional complexity. We propose two methods. In tunnel networks, this selection is done at the level of a hidden unit, and in budding perceptrons, this is done at the level of a network layer; updating this control parameter introduces either another hidden unit or layer. We show the effectiveness of our methods on the synthetic two-spiral data and on three real data sets of MNIST, MIRFLICKR, and CIFAR, where we see that our proposed methods, with the same set of hyperparameters, can correctly adjust the network complexity to the task complexity.
AbstractList Traditionally, deep learning algorithms update the network weights, whereas the network architecture is chosen manually using a process of trial and error. In this paper, we propose two novel approaches that automatically update the network structure while also learning its weights. The novelty of our approach lies in our parameterization, where the depth, or additional complexity, is encapsulated continuously in the parameter space through control parameters that add additional complexity. We propose two methods. In tunnel networks, this selection is done at the level of a hidden unit, and in budding perceptrons, this is done at the level of a network layer; updating this control parameter introduces either another hidden unit or layer. We show the effectiveness of our methods on the synthetic two-spiral data and on three real data sets of MNIST, MIRFLICKR, and CIFAR, where we see that our proposed methods, with the same set of hyperparameters, can correctly adjust the network complexity to the task complexity.
Traditionally, deep learning algorithms update the network weights, whereas the network architecture is chosen manually using a process of trial and error. In this paper, we propose two novel approaches that automatically update the network structure while also learning its weights. The novelty of our approach lies in our parameterization, where the depth, or additional complexity, is encapsulated continuously in the parameter space through control parameters that add additional complexity. We propose two methods. In tunnel networks, this selection is done at the level of a hidden unit, and in budding perceptrons, this is done at the level of a network layer; updating this control parameter introduces either another hidden unit or layer. We show the effectiveness of our methods on the synthetic two-spiral data and on three real data sets of MNIST, MIRFLICKR, and CIFAR, where we see that our proposed methods, with the same set of hyperparameters, can correctly adjust the network complexity to the task complexity.Traditionally, deep learning algorithms update the network weights, whereas the network architecture is chosen manually using a process of trial and error. In this paper, we propose two novel approaches that automatically update the network structure while also learning its weights. The novelty of our approach lies in our parameterization, where the depth, or additional complexity, is encapsulated continuously in the parameter space through control parameters that add additional complexity. We propose two methods. In tunnel networks, this selection is done at the level of a hidden unit, and in budding perceptrons, this is done at the level of a network layer; updating this control parameter introduces either another hidden unit or layer. We show the effectiveness of our methods on the synthetic two-spiral data and on three real data sets of MNIST, MIRFLICKR, and CIFAR, where we see that our proposed methods, with the same set of hyperparameters, can correctly adjust the network complexity to the task complexity.
Author Alpaydin, Ethem
Irsoy, Ozan
Author_xml – sequence: 1
  givenname: Ozan
  orcidid: 0000-0002-7123-8361
  surname: Irsoy
  fullname: Irsoy, Ozan
  email: oirsoy@bloomberg.net
  organization: Bloomberg LP, New York, NY, USA
– sequence: 2
  givenname: Ethem
  orcidid: 0000-0001-7506-0321
  surname: Alpaydin
  fullname: Alpaydin, Ethem
  email: alpaydin@boun.edu.tr
  organization: Department of Computer Engineering, Boğaziçi University, Istanbul, Turkey
BackLink https://www.ncbi.nlm.nih.gov/pubmed/31247565$$D View this record in MEDLINE/PubMed
BookMark eNp9kEtLAzEQx4NUbK39AgpS8OJl17w2j6PUJ5R6sIK3sI8JbN3u1mRX6bc3ta2HHsxlMvD7DzO_U9SrmxoQOic4JgTrm_lsNn2NKSY6ppooSpMjNKBE0IgypXp_f_neRyPvFzg8gRPB9QnqM0K5TEQyQPGkqduy7prOV-txaHzrurwtv2B8B7Aaz6BzaRVK-924D3-Gjm1aeRjt6hC9PdzPJ0_R9OXxeXI7jXKWkDYSKSssV5m12nKihSXUUqsJYYoXBETCKM6Bg-BCgdJCqIxSEFlipZJUFmyIrrdzV6757MC3Zln6HKoqrSGsasK5WITjCAvo1QG6aDpXh-1MMCGxxInUgbrcUV22hMKsXLlM3drsTQRAbYHcNd47sCYv27Qtgx6XlpUh2Gy8m1_vZuPd7LyHKD2I7qf_G7rYhkoA-AsoyTlXiv0A6NeLiQ
CODEN ITNNAL
CitedBy_id crossref_primary_10_1016_j_neunet_2021_08_034
crossref_primary_10_3390_a14050158
crossref_primary_10_1109_ACCESS_2024_3376525
crossref_primary_10_1145_3696003
crossref_primary_10_1016_j_bspc_2022_104185
crossref_primary_10_1177_01423312211055569
crossref_primary_10_1016_j_engappai_2022_105734
crossref_primary_10_4316_AECE_2020_02005
crossref_primary_10_3390_s22155786
Cites_doi 10.1109/72.279181
10.1109/TNN.2002.804317
10.1109/CVPR.2016.308
10.1109/ICCV.2017.298
10.1561/2200000006
10.1109/72.838999
10.1016/0893-6080(94)90058-2
10.1162/089976601317098565
10.1007/978-3-030-01246-5_1
10.1016/j.neucom.2003.10.009
10.1145/2733373.2806216
10.1109/CVPR.2016.90
10.1080/09540098908915647
10.1109/72.248452
10.1109/TNN.2004.836241
10.1109/ICMLA.2017.0-160
10.1142/S0218001409007132
10.1162/neco.1997.9.8.1735
10.1109/72.572092
10.1109/ICPR.2014.616
10.1109/72.774273
ContentType Journal Article
Copyright Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020
Copyright_xml – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2020
DBID 97E
RIA
RIE
AAYXX
CITATION
NPM
7QF
7QO
7QP
7QQ
7QR
7SC
7SE
7SP
7SR
7TA
7TB
7TK
7U5
8BQ
8FD
F28
FR3
H8D
JG9
JQ2
KR7
L7M
L~C
L~D
P64
7X8
DOI 10.1109/TNNLS.2019.2918225
DatabaseName IEEE All-Society Periodicals Package (ASPP) 2005–Present
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE Electronic Library (IEL)
CrossRef
PubMed
Aluminium Industry Abstracts
Biotechnology Research Abstracts
Calcium & Calcified Tissue Abstracts
Ceramic Abstracts
Chemoreception Abstracts
Computer and Information Systems Abstracts
Corrosion Abstracts
Electronics & Communications Abstracts
Engineered Materials Abstracts
Materials Business File
Mechanical & Transportation Engineering Abstracts
Neurosciences Abstracts
Solid State and Superconductivity Abstracts
METADEX
Technology Research Database
ANTE: Abstracts in New Technology & Engineering
Engineering Research Database
Aerospace Database
Materials Research Database
ProQuest Computer Science Collection
Civil Engineering Abstracts
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
Biotechnology and BioEngineering Abstracts
MEDLINE - Academic
DatabaseTitle CrossRef
PubMed
Materials Research Database
Technology Research Database
Computer and Information Systems Abstracts – Academic
Mechanical & Transportation Engineering Abstracts
ProQuest Computer Science Collection
Computer and Information Systems Abstracts
Materials Business File
Aerospace Database
Engineered Materials Abstracts
Biotechnology Research Abstracts
Chemoreception Abstracts
Advanced Technologies Database with Aerospace
ANTE: Abstracts in New Technology & Engineering
Civil Engineering Abstracts
Aluminium Industry Abstracts
Electronics & Communications Abstracts
Ceramic Abstracts
Neurosciences Abstracts
METADEX
Biotechnology and BioEngineering Abstracts
Computer and Information Systems Abstracts Professional
Solid State and Superconductivity Abstracts
Engineering Research Database
Calcium & Calcified Tissue Abstracts
Corrosion Abstracts
MEDLINE - Academic
DatabaseTitleList
MEDLINE - Academic
PubMed
Materials Research Database
Database_xml – sequence: 1
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 2
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
– sequence: 3
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Computer Science
EISSN 2162-2388
EndPage 1133
ExternalDocumentID 31247565
10_1109_TNNLS_2019_2918225
8744488
Genre orig-research
Journal Article
GroupedDBID 0R~
4.4
5VS
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABQJQ
ABVLG
ACIWK
ACPRK
AENEX
AFRAH
AGQYO
AGSQL
AHBIQ
AKJIK
AKQYR
ALMA_UNASSIGNED_HOLDINGS
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
EBS
EJD
IFIPE
IPLJI
JAVBF
M43
MS~
O9-
OCL
PQQKQ
RIA
RIE
RNS
AAYXX
CITATION
NPM
RIG
7QF
7QO
7QP
7QQ
7QR
7SC
7SE
7SP
7SR
7TA
7TB
7TK
7U5
8BQ
8FD
F28
FR3
H8D
JG9
JQ2
KR7
L7M
L~C
L~D
P64
7X8
ID FETCH-LOGICAL-c351t-6a3df48bff9f4196f12f2f911384d1e65320ce4e6468e89668b22e6b5f78727d3
IEDL.DBID RIE
ISICitedReferencesCount 13
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000525351800005&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 2162-237X
2162-2388
IngestDate Thu Oct 02 09:54:57 EDT 2025
Mon Jun 30 03:52:13 EDT 2025
Thu Jan 02 22:59:03 EST 2025
Tue Nov 18 21:49:20 EST 2025
Sat Nov 29 01:40:03 EST 2025
Wed Aug 27 02:42:21 EDT 2025
IsPeerReviewed false
IsScholarly true
Issue 4
Language English
License https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html
https://doi.org/10.15223/policy-029
https://doi.org/10.15223/policy-037
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c351t-6a3df48bff9f4196f12f2f911384d1e65320ce4e6468e89668b22e6b5f78727d3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ORCID 0000-0001-7506-0321
0000-0002-7123-8361
PMID 31247565
PQID 2387070579
PQPubID 85436
PageCount 10
ParticipantIDs proquest_journals_2387070579
ieee_primary_8744488
proquest_miscellaneous_2250623713
pubmed_primary_31247565
crossref_citationtrail_10_1109_TNNLS_2019_2918225
crossref_primary_10_1109_TNNLS_2019_2918225
PublicationCentury 2000
PublicationDate 2020-04-01
PublicationDateYYYYMMDD 2020-04-01
PublicationDate_xml – month: 04
  year: 2020
  text: 2020-04-01
  day: 01
PublicationDecade 2020
PublicationPlace United States
PublicationPlace_xml – name: United States
– name: Piscataway
PublicationTitle IEEE transaction on neural networks and learning systems
PublicationTitleAbbrev TNNLS
PublicationTitleAlternate IEEE Trans Neural Netw Learn Syst
PublicationYear 2020
Publisher IEEE
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml – name: IEEE
– name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References ref12
ref37
ref2
ref1
ref17
ref16
ref19
ref18
huiskes (ref34) 2008
glorot (ref32) 2011; 15
bergstra (ref13) 2012; 13
ref24
ref23
ref25
ref20
ref22
ref21
cortes (ref10) 2016
ref28
hinton (ref33) 2012
ref27
srivastava (ref11) 2015
krizhevsky (ref38) 2012
ref29
zoph (ref9) 2016
kingma (ref31) 2014
józefowicz (ref8) 2015
ref7
ref4
ref3
ref6
srivastava (ref35) 2014; 15
fahlman (ref15) 1990
lang (ref30) 1988
krizhevsky (ref36) 2009; 1
snoek (ref14) 2015
chung (ref5) 2014
huang (ref26) 2017
References_xml – ident: ref3
  doi: 10.1109/72.279181
– ident: ref24
  doi: 10.1109/TNN.2002.804317
– volume: 15
  start-page: 315
  year: 2011
  ident: ref32
  article-title: Deep sparse rectifier networks
  publication-title: J Mach Learn Res
– volume: 13
  start-page: 281
  year: 2012
  ident: ref13
  article-title: Random search for hyper-parameter optimization
  publication-title: J Mach Learn Res
– ident: ref37
  doi: 10.1109/CVPR.2016.308
– ident: ref25
  doi: 10.1109/ICCV.2017.298
– start-page: 1097
  year: 2012
  ident: ref38
  article-title: ImageNet classification with deep convolutional neural networks
  publication-title: Proc Adv Neural Inf Process Syst
– ident: ref1
  doi: 10.1561/2200000006
– ident: ref20
  doi: 10.1109/72.838999
– start-page: 39
  year: 2008
  ident: ref34
  article-title: The MIR Flickr retrieval evaluation
  publication-title: Proc Int Conf Multimedia Inf Retr
– year: 2014
  ident: ref31
  article-title: Adam: A method for stochastic optimization
  publication-title: arXiv 1412 6980
– year: 2014
  ident: ref5
  article-title: Empirical evaluation of gated recurrent neural networks on sequence modeling
  publication-title: arXiv 1412 3555
– start-page: 52
  year: 1988
  ident: ref30
  article-title: Learning to tell two spirals apart
  publication-title: Proc Connectionist Models Summer School
– ident: ref21
  doi: 10.1016/0893-6080(94)90058-2
– volume: 1
  start-page: 7
  year: 2009
  ident: ref36
  article-title: Learning multiple layers of features from tiny images
– ident: ref16
  doi: 10.1162/089976601317098565
– year: 2015
  ident: ref11
  article-title: Highway networks
  publication-title: arXiv 1505 00387
– ident: ref28
  doi: 10.1007/978-3-030-01246-5_1
– ident: ref19
  doi: 10.1016/j.neucom.2003.10.009
– ident: ref2
  doi: 10.1145/2733373.2806216
– ident: ref29
  doi: 10.1109/CVPR.2016.90
– ident: ref6
  doi: 10.1080/09540098908915647
– start-page: 2342
  year: 2015
  ident: ref8
  article-title: An empirical exploration of recurrent network architectures
  publication-title: Proc 32nd Int Conf Mach Learn (ICML)
– year: 2012
  ident: ref33
  article-title: Improving neural networks by preventing co-adaptation of feature detectors
  publication-title: arXiv 1207 0580
– ident: ref7
  doi: 10.1109/72.248452
– ident: ref22
  doi: 10.1109/TNN.2004.836241
– volume: 15
  start-page: 2949
  year: 2014
  ident: ref35
  article-title: Multimodal learning with deep Boltzmann machines
  publication-title: J Mach Learn Res
– ident: ref27
  doi: 10.1109/ICMLA.2017.0-160
– ident: ref23
  doi: 10.1142/S0218001409007132
– year: 2016
  ident: ref10
  article-title: AdaNet: Adaptive structural learning of artificial neural networks
  publication-title: arXiv 1607 01097
– ident: ref4
  doi: 10.1162/neco.1997.9.8.1735
– start-page: 2171
  year: 2015
  ident: ref14
  article-title: Scalable Bayesian optimization using deep neural networks
  publication-title: Proc Int Conf Mach Learn
– ident: ref17
  doi: 10.1109/72.572092
– ident: ref12
  doi: 10.1109/ICPR.2014.616
– start-page: 524
  year: 1990
  ident: ref15
  article-title: The cascade-correlation learning architecture
  publication-title: Proc Adv Neural Inf Process Syst
– ident: ref18
  doi: 10.1109/72.774273
– year: 2017
  ident: ref26
  article-title: Data-driven sparse structure selection for deep neural networks
  publication-title: arXiv 1707 01213
– year: 2016
  ident: ref9
  article-title: Neural architecture search with reinforcement learning
  publication-title: arXiv 1611 01578
SSID ssj0000605649
Score 2.414827
Snippet Traditionally, deep learning algorithms update the network weights, whereas the network architecture is chosen manually using a process of trial and error. In...
SourceID proquest
pubmed
crossref
ieee
SourceType Aggregation Database
Index Database
Enrichment Source
Publisher
StartPage 1124
SubjectTerms Algorithms
Artificial neural networks
Complexity theory
Computer architecture
Constructive learning
Deep learning
Learning algorithms
Machine learning
Network architecture
Neural networks
Parameterization
Parameters
Road transportation
Task complexity
Training
Title Continuously Constructive Deep Neural Networks
URI https://ieeexplore.ieee.org/document/8744488
https://www.ncbi.nlm.nih.gov/pubmed/31247565
https://www.proquest.com/docview/2387070579
https://www.proquest.com/docview/2250623713
Volume 31
WOSCitedRecordID wos000525351800005&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVIEE
  databaseName: IEEE Electronic Library (IEL)
  customDbUrl:
  eissn: 2162-2388
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0000605649
  issn: 2162-237X
  databaseCode: RIE
  dateStart: 20120101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8QwEB7cxYMX34_qulTwpl3bNM3jKOriYSmCK-yt9DEBYemKq4L_3knaLR5U8FIKTdIwmUm-SSbzAZxXiuU64SLQLOcB1_QoFJKXEvKCHBIToXGsJROZpmo20w9rcNndhUFEF3yGI_vqzvKrRflut8qubKp2Urge9KQUzV2tbj8lJFwuHNplkWABi-VsdUcm1FfTNJ082kAuPWKaILVlxv62Djlild8xpltrxlv_6-U2bLaY0r9ulGAH1rDeha0VX4Pfmu8ejGwqquf6nZz9-advqTqb5LEf6N8ivvg2UQe1kzaR4ct9eBrfTW_ug5YvISjjJHoLRB5XhqvCGG04WZaJmGGGZrNY8SpCYTkgSuQouFCoyM9RBWMoisSQ1TJZxQfQrxc1HoGfa2mqOBYh0vJflFHOVKWEMLaGKMvCg2glvaxsk4lbTot55pyKUGdO4pmVeNZK3IOLrs5Lk0rjz9J7VrRdyVaqHgxWg5S11rbMCHZImroSqT046z6TndjDj7xGEmtGTYYE9cgn9-CwGdyu7ZhAjiRke_zzP09gg1kv28XrDKBPY4OnsF5-vD0vX4ekjDM1dMr4BY902FM
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1ZS8QwEB68QF-8j3pW8E27tmmaJo_igeJaBFfYt9BjAgtSxVXBf-8k7RYfVPClFJqk5Usm_SaZzAdwVEmWq4SLQLGcB1zRpZBIXkrIC3JITITGqZb00yyTw6G6n4KT7iwMIrrgM-zZW7eXXz2X73ap7NSmaqcBNw2zCecsbE5rdSsqITFz4fguiwQLWJwOJ6dkQnU6yLL-gw3lUj2miFRbbexvfyInrfI7y3R_m6ul_33nMiy2rNI_a4bBCkxhvQpLE8UGvzXgNejZZFSj-p3c_adP34p1NuljP9C_QHzxbaoOaidrYsPH6_B4dTk4vw5axYSgjJPoLRB5XBkuC2OU4WRbJmKGGZrPYsmrCIVVgSiRo-BCoiRPRxaMoSgSQ3bL0iregJn6ucYt8HOVmiqORYhEAIoyypmspBDG1hBlWXgQTdDTZZtO3KpaPGnnVoRKO8S1RVy3iHtw3NV5aZJp_Fl6zULblWxR9WB30km6tbexJuKR0uSVpMqDw-4xWYrd_shrJFg1NRkS2SOv3IPNpnO7tmOiOSlx2-2f33kA89eDu77u32S3O7DArM_tond2YYb6Cfdgrvx4G41f992Q_ALffNqy
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=Continuously+Constructive+Deep+Neural+Networks&rft.jtitle=IEEE+transaction+on+neural+networks+and+learning+systems&rft.au=Irsoy%2C+Ozan&rft.au=Alpaydin%2C+Ethem&rft.date=2020-04-01&rft.issn=2162-237X&rft.eissn=2162-2388&rft.volume=31&rft.issue=4&rft.spage=1124&rft.epage=1133&rft_id=info:doi/10.1109%2FTNNLS.2019.2918225&rft.externalDBID=n%2Fa&rft.externalDocID=10_1109_TNNLS_2019_2918225
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2162-237X&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2162-237X&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2162-237X&client=summon