Bit-serial convolution with prediction threshold for convolutional neural networks Electrical Engineering Subject Index: EL7 Signal Processing

To reduce the implementation complexity and power consumption of the convolution operation in a convolutional neural network (CNN), this work proposes a new convolution method using the serial input and prediction threshold. To confirm the benefits of the proposed method, we use the original paramet...

Full description

Saved in:
Bibliographic Details
Published inJournal of the Chinese Institute of Engineers Vol. 45; no. 3; pp. 266 - 272
Main Authors Hsiao, Jen-Hao, Chin, Wen-Long, Wu, Yu-Feng, Chang, Deng-Kai
Format Journal Article
LanguageEnglish
Published Taylor & Francis 03.04.2022
Subjects
Online AccessGet full text
ISSN0253-3839
2158-7299
DOI10.1080/02533839.2022.2034050

Cover

Abstract To reduce the implementation complexity and power consumption of the convolution operation in a convolutional neural network (CNN), this work proposes a new convolution method using the serial input and prediction threshold. To confirm the benefits of the proposed method, we use the original parameters, i.e. kernel weights and biases, of the popular AlexNet to verify the proposed algorithm and then implement its digital circuit. According to implementation data and comparison with traditional convolution using bit-parallel input, the implementation gain in terms of the throughput/power/area of the serial convolution is 7.57 times that of parallel convolution.
AbstractList To reduce the implementation complexity and power consumption of the convolution operation in a convolutional neural network (CNN), this work proposes a new convolution method using the serial input and prediction threshold. To confirm the benefits of the proposed method, we use the original parameters, i.e. kernel weights and biases, of the popular AlexNet to verify the proposed algorithm and then implement its digital circuit. According to implementation data and comparison with traditional convolution using bit-parallel input, the implementation gain in terms of the throughput/power/area of the serial convolution is 7.57 times that of parallel convolution.
Author Hsiao, Jen-Hao
Wu, Yu-Feng
Chin, Wen-Long
Chang, Deng-Kai
Author_xml – sequence: 1
  givenname: Jen-Hao
  surname: Hsiao
  fullname: Hsiao, Jen-Hao
  organization: National Cheng Kung University
– sequence: 2
  givenname: Wen-Long
  surname: Chin
  fullname: Chin, Wen-Long
  email: wlchin@mail.ncku.edu.tw
  organization: National Cheng Kung University
– sequence: 3
  givenname: Yu-Feng
  surname: Wu
  fullname: Wu, Yu-Feng
  organization: National Cheng Kung University
– sequence: 4
  givenname: Deng-Kai
  surname: Chang
  fullname: Chang, Deng-Kai
  organization: National Cheng Kung University
BookMark eNp9kF9LwzAUxYNMcJt-BKFfoDN_mjZ9U4f_YCCIPoc0vaHRLhlJ5ti3t90m-OTLPVz4ncO9Z4YmzjtA6JrgBcEC32DKGROsXlBM6TBYgTk-Q1NKuMgrWtcTNB2ZfIQu0CzGT4xpySmZord7m_IIwao-0959-36brHfZzqYu2wRorT7sqQsQO9-3mfHhLzn4HGzDQdLOh694ic6N6iNcnXSOPh4f3pfP-er16WV5t8o15SLlULS1KEtVEVCsAa6VobyqoCINx6C0ok0hiKlLUjWtgYaagaHQaq5qwYGwOeLHXB18jAGM3AS7VmEvCZZjMfK3GDkWI0_FDL7bo8-64Ze1Go7uW5nUvvfBBOW0jZL9H_EDxFdvKg
Cites_doi 10.1109/MNET.011.2000096
10.1049/cmu2.12129
10.1109/MICRO.2016.7783722
10.1109/ICASSP.2017.7953288
10.1145/3065386
10.1109/ICIS.2016.7550778
10.1007/s11263-015-0816-y
ContentType Journal Article
Copyright 2022 The Chinese Institute of Engineers 2022
Copyright_xml – notice: 2022 The Chinese Institute of Engineers 2022
DBID AAYXX
CITATION
DOI 10.1080/02533839.2022.2034050
DatabaseName CrossRef
DatabaseTitle CrossRef
DatabaseTitleList
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
EISSN 2158-7299
EndPage 272
ExternalDocumentID 10_1080_02533839_2022_2034050
2034050
Genre Research Article
GroupedDBID -~X
.7F
.QJ
0BK
0R~
30N
4.4
5GY
AAENE
AAGDL
AAHIA
AAJMT
AALDU
AAMIU
AAPUL
AAQRR
ABCCY
ABFIM
ABLIJ
ABPAQ
ABPEM
ABTAI
ABXUL
ABXYU
ACGFS
ACTIO
ADCVX
ADGTB
AEISY
AENEX
AEOZL
AFRVT
AGDLA
AGMYJ
AHDZW
AIJEM
AIYEW
AKBVH
AKOOK
ALMA_UNASSIGNED_HOLDINGS
ALQZU
AQRUH
AQTUD
AVBZW
BLEHA
CCCUG
DGEBU
DKSSO
EBS
E~A
E~B
FRP
GTTXZ
H13
HF~
HZ~
H~P
J.P
KYCEM
LJTGL
M4Z
NA5
NX~
O9-
RIG
RNANH
ROSJB
RTWRZ
S-T
SNACF
TASJS
TBQAZ
TDBHL
TEN
TFL
TFT
TFW
TTHFI
TUROJ
UT5
UU3
ZGOLN
~S~
AAYXX
CITATION
ID FETCH-LOGICAL-c258t-e4d9866a71ea3be5caf2577e71b50eaca2b481f9617bdfeb2fe5c2edc5a985e13
ISSN 0253-3839
IngestDate Wed Oct 01 03:54:15 EDT 2025
Mon Oct 20 23:50:04 EDT 2025
IsPeerReviewed true
IsScholarly true
Issue 3
Language English
LinkModel OpenURL
MergedId FETCHMERGED-LOGICAL-c258t-e4d9866a71ea3be5caf2577e71b50eaca2b481f9617bdfeb2fe5c2edc5a985e13
PageCount 7
ParticipantIDs informaworld_taylorfrancis_310_1080_02533839_2022_2034050
crossref_primary_10_1080_02533839_2022_2034050
ProviderPackageCode CITATION
AAYXX
PublicationCentury 2000
PublicationDate 2022-04-03
PublicationDateYYYYMMDD 2022-04-03
PublicationDate_xml – month: 04
  year: 2022
  text: 2022-04-03
  day: 03
PublicationDecade 2020
PublicationTitle Journal of the Chinese Institute of Engineers
PublicationYear 2022
Publisher Taylor & Francis
Publisher_xml – name: Taylor & Francis
References Courbariaux M. (cit0003) 2015
cit0001
cit0010
cit0008
cit0009
cit0006
cit0007
cit0004
cit0005
cit0002
References_xml – start-page: 3123
  volume-title: Proceedings of the 28th International Conference on Neural Information Processing Systems
  year: 2015
  ident: cit0003
– ident: cit0001
  doi: 10.1109/MNET.011.2000096
– ident: cit0002
  doi: 10.1049/cmu2.12129
– ident: cit0005
– ident: cit0004
– ident: cit0006
  doi: 10.1109/MICRO.2016.7783722
– ident: cit0008
  doi: 10.1109/ICASSP.2017.7953288
– ident: cit0007
  doi: 10.1145/3065386
– ident: cit0010
  doi: 10.1109/ICIS.2016.7550778
– ident: cit0009
  doi: 10.1007/s11263-015-0816-y
SSID ssj0026521
ssj0001654702
Score 2.2333798
Snippet To reduce the implementation complexity and power consumption of the convolution operation in a convolutional neural network (CNN), this work proposes a new...
SourceID crossref
informaworld
SourceType Index Database
Publisher
StartPage 266
SubjectTerms Convolutional neural network
hardware accelerator
serial input
Subtitle Electrical Engineering Subject Index: EL7 Signal Processing
Title Bit-serial convolution with prediction threshold for convolutional neural networks
URI https://www.tandfonline.com/doi/abs/10.1080/02533839.2022.2034050
Volume 45
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVLSH
  databaseName: aylor and Francis Online
  customDbUrl:
  mediaType: online
  eissn: 2158-7299
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0026521
  issn: 0253-3839
  databaseCode: AHDZW
  dateStart: 19970101
  isFulltext: true
  providerName: Library Specific Holdings
– providerCode: PRVAWR
  databaseName: Taylor & Francis Science and Technology Library-DRAA
  customDbUrl:
  eissn: 2158-7299
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0026521
  issn: 0253-3839
  databaseCode: 30N
  dateStart: 19970101
  isFulltext: true
  titleUrlDefault: http://www.tandfonline.com/page/title-lists
  providerName: Taylor & Francis
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV3Pb9MwFLZKd4EDGgzEYCAfuLpKnNhxjoNtioDtMHXaxCWyUwf10k5reuGP4G_mPdtJXG0gflzc1o372ve--j07z98j5L3kiWrb3DKdaMHyxiqmG2NZqcAdIHtKo_E08vmFrK7yTzfiZjL5EWUtbTsza74_eK7kX6wKfWBXPCX7F5YdPhQ64DnYF1qwMLR_ZOMPy455QS57PMjye6u3d3gLxr3uwF4bvM3kcgqjK2Ec8lm6B5cNvvlFrIrRKVbathu7m17Q0xkOkXm1Wep1SJthlV6P6QOerOAaur-sg7tEZ7B1TmDLQLHfxovDLvYJdLLPehnvTcCyFlNasgFN83tlQqLZjYuMwfLYz5fW9UH4obC2bhlPz55tMsAwi-daKSO3zX0FoHseIaRQgjQUNsNvCU0GcWoyusAhMTG884jscfASyZTsHVcnX6_HXTus1ZwMBGVcCneyb_gx_RkxZG9_SORO9LPDjRtFNfN98jSYmB57bD0jE7t6Tp5EJJUH5HJEGY2wQxFldEQZHVBGQRzdQRn1KKM9yl6Qq7PT-ceKhUocrOFCdczmi1JJqYvU6sxY-AO3MNUXtkiNSMB1a25ylbYlhMNm0VrDW7iG20UjdKmETbOXZLpar-wrQqWShRU8EQZW_uBdyrSEkVlbQFzaLExySGa9hupbT7hSpz2PbVBpjSqtg0oPSRnrse4c6FqPtzr77djX_zH2DXk8Qv6ITLu7rX0LAWpn3gXU_ARZtIyO
linkProvider Library Specific Holdings
linkToHtml http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV07T8MwELZQGYCBN6I8M7C6JHGcOCMgqgJtB9RK3SLbsSWEVKo2Xfj13DkJpEiwdIqi-CzZvtzLd98RchOHvrA2MlT6ktNIG0GlVoamAtQBoqdoidXIg2HcG0fPEz5p1MJgWiX60LYEinCyGn9uDEbXKXG3oKfRs8I6kxCLqRhYHeC2b3Iw9rGLAfOHP3EW7K7rf0NKhTF3tVg4BcU56qqev6Zd0VcraKYNPdTdI7peQZl-8t5ZFqqjP3-BO663xH2yW5mp3l3JVwdkw0wPyU4DvPCIvN6_FbRkYA9z1yse9jCy683meAHk3gvglgVecnmwqOZIoEM0TfdwueiLYzLuPo4eerTq0EB1yEVBTZSnIo5lEhjJlIGDtSACEpMEivsg0mWoIhHYFMwklVtw4i2MCU2uuUwFNwE7Ia3px9ScEi8WcWJ46HMFHiFInTRIgZLZBOwVnSu_TTr1OWSzEogjC2p802qzMtysrNqsNkmbp5UVLgJiy3YlGfuX9mwN2muy1RsN-ln_afhyTrbxk8v0YRekVcyX5hKMmEJdOS79AsEt49E
linkToPdf http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1LS8QwEB5kBdGDb3F99uA1a19J06OvZX0tIi54K0mbgAjrstu9-OudSVvdFfSyp1KaCSSZziP55gvAmQh9aW1smPIVZ3FuJFO5NiyV6A6IPSVXVI382Be9QXz3yhs04aSGVVIObSuiCGer6eceFbZBxJ2jm6bEispMQqqlijDowKx9WdCpGFVx-P2fbRa6XNf_ZpQKBXelWNQFoz6aop6_up1zV3NkpjNuqLsBuhlAhT5570xL3ck_f3E7LjTCTVivg1TvotKqLVgyw21Ym6Eu3IHny7eSVerrEXK91mCP9nW90ZiOf9x7iboyoSMuD8c02xLliEvTPRwSfbILg-7Ny1WP1fczsDzksmQmLlIphEoCoyJtcFktGoDEJIHmPhp0FepYBjbFIEkXFlN4i21CU-RcpZKbINqD1vBjaPbBE1Ikhoc-15gPos1JgxQlI5tgtJIX2m9Dp1mGbFTRcGRBw25aT1ZGk5XVk9WGdHaxstLtf9jqspIs-lf2YAHZU1h5uu5mD7f9-0NYpS8O5hMdQascT80xRjClPnE6-gUFkeJ1
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=Bit-serial+convolution+with+prediction+threshold+for+convolutional+neural+networks&rft.jtitle=Journal+of+the+Chinese+Institute+of+Engineers&rft.au=Hsiao%2C+Jen-Hao&rft.au=Chin%2C+Wen-Long&rft.au=Wu%2C+Yu-Feng&rft.au=Chang%2C+Deng-Kai&rft.date=2022-04-03&rft.pub=Taylor+%26+Francis&rft.issn=0253-3839&rft.eissn=2158-7299&rft.volume=45&rft.issue=3&rft.spage=266&rft.epage=272&rft_id=info:doi/10.1080%2F02533839.2022.2034050&rft.externalDocID=2034050
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0253-3839&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0253-3839&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0253-3839&client=summon