Mixing floating- and fixed-point formats for neural network learning on neuroprocessors

We examine the efficient implementation of back-propagation (BP) type algorithms on TO [3], a vector processor with a fixed-point engine, designed for neural network simulation. Using Matrix Back Propagation (MBP) [2]we achieve an asymptotically optimal performance on TO (about 0.8 GOPS) for both fo...

Full description

Saved in:
Bibliographic Details
Published inMicroprocessing and microprogramming Vol. 41; no. 10; pp. 757 - 769
Main Authors Anguita, Davide, Gomes, Benedict A.
Format Journal Article
LanguageEnglish
Published Elsevier B.V 01.06.1996
Subjects
Online AccessGet full text
ISSN0165-6074
DOI10.1016/0165-6074(96)00012-9

Cover

Abstract We examine the efficient implementation of back-propagation (BP) type algorithms on TO [3], a vector processor with a fixed-point engine, designed for neural network simulation. Using Matrix Back Propagation (MBP) [2]we achieve an asymptotically optimal performance on TO (about 0.8 GOPS) for both forward and backward phases, which is not possible with the standard on-line BP algorithm. We use a mixture of fixed- and floating-point operations in order to guarantee both high efficiency and fast convergence. Though the most expensive computations are implemented in fixed-point, we achieve a rate of convergence that is comparable to the floating-point version. The time taken for conversion between fixed- and floating-point is also shown to be reasonably low.
AbstractList We examine the efficient implementation of back-propagation (BP) type algorithms on TO [3], a vector processor with a fixed-point engine, designed for neural network simulation. Using Matrix Back Propagation (MBP) [2]we achieve an asymptotically optimal performance on TO (about 0.8 GOPS) for both forward and backward phases, which is not possible with the standard on-line BP algorithm. We use a mixture of fixed- and floating-point operations in order to guarantee both high efficiency and fast convergence. Though the most expensive computations are implemented in fixed-point, we achieve a rate of convergence that is comparable to the floating-point version. The time taken for conversion between fixed- and floating-point is also shown to be reasonably low.
We examine the efficient implementation of back-propagation (BP) type algorithms on T0 [3], a vector processor with a fixed-point engine, designed for neural network simulation. Using Matrix Back Propagation (MBP) [2] we achieve an asymptotically optimal performance on T0 (about 0.8 GOPS) for both forward and backward phases, which is not possible with the standard on-line BP algorithm. We use a mixture of fixed- and floating-point operations in order to guarantee both high efficiency and fast convergence. Though the most expensive computations are implemented in fixed-point, we achieve a rate of convergence that is comparable to the floating-point version. The time taken for conversion between fixed- and floating-point is also shown to be reasonably low.
Author Anguita, Davide
Gomes, Benedict A.
Author_xml – sequence: 1
  givenname: Davide
  surname: Anguita
  fullname: Anguita, Davide
  email: anguita@dibe.unige.it
  organization: Univ. of Genova, D.I.B.E., Via Opera Pia 11a, 16145 Genova, Italy
– sequence: 2
  givenname: Benedict A.
  surname: Gomes
  fullname: Gomes, Benedict A.
  email: gomes@icsi.berkeley.edu
  organization: Int. Comp. Science Inst., 1947 Center St., Berkeley, CA, USA
BookMark eNqFkLlOAzEURV0EiQT4A4qpEBQD3sYTUyChiE0KogFRWh4vyDCxg-1A-Hs8BFFQQPF033ae5TsBIx-8AWAfwWMEETsp0dQMtvSQsyMIIcI1H4HxT3sbTFJ6LgPK22YMHm_d2vmnyvZB5pLUlfS6sm5tdL0MzufKhriQOQ1aebOKsi-S30N8qXojox_o4L9GYRmDMimFmHbBlpV9MnvfugMeLi_uZ9f1_O7qZnY-rxUhTa4JnVLSWAKxlUY2U6wQQ4hy0rUaY9VR2XVYN5B3U2IRJEzrVllKu9JrlWrJDjjY3C1Pv65MymLhkjJ9L70JqyQww5Azxsri6WZRxZBSNFYol8uXg89Rul4gKAb_xGCUGIwSvBSDf4IXmP6Cl9EtZPz4DzvbYKY48OZMFEk545XRLhqVhQ7u7wOfwmKMRA
CitedBy_id crossref_primary_10_1016_S1383_7621_99_00017_X
crossref_primary_10_1016_j_neunet_2014_09_003
crossref_primary_10_1016_j_neucom_2007_12_006
crossref_primary_10_1631_jzus_C1000137
Cites_doi 10.1109/72.143374
10.1007/BF00332914
10.1109/78.215327
10.1016/0893-6080(90)90006-7
10.1162/neco.1992.4.6.835
10.1109/72.129414
10.1162/neco.1990.2.3.363
10.1007/BF02551274
10.1002/aic.690370209
10.1142/S0129065793000250
10.1109/72.207614
10.1016/0925-2312(94)90034-5
10.1109/72.217180
10.1109/72.286885
ContentType Journal Article
Copyright 1996
Copyright_xml – notice: 1996
DBID AAYXX
CITATION
7SC
8FD
JQ2
L7M
L~C
L~D
DOI 10.1016/0165-6074(96)00012-9
DatabaseName CrossRef
Computer and Information Systems Abstracts
Technology Research Database
ProQuest Computer Science Collection
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
DatabaseTitle CrossRef
Computer and Information Systems Abstracts
Technology Research Database
Computer and Information Systems Abstracts – Academic
Advanced Technologies Database with Aerospace
ProQuest Computer Science Collection
Computer and Information Systems Abstracts Professional
DatabaseTitleList
Computer and Information Systems Abstracts
DeliveryMethod fulltext_linktorsrc
Discipline Computer Science
EndPage 769
ExternalDocumentID 10_1016_0165_6074_96_00012_9
0165607496000129
GroupedDBID --K
--M
-~X
.~1
0R~
123
1B1
1~.
5VS
8P~
AACTN
AAEDT
AAEDW
AAIKJ
AAKOC
AALRI
AAOAW
AAQFI
AAQXK
AAXUO
AAYFN
ABBOA
ABMAC
ABXDB
ABYKQ
ACDAQ
ACGFS
ACNNM
ACRLP
ADBBV
ADEZE
ADJOM
ADMUD
AEKER
AFKWA
AFTJW
AGHFR
AGUBO
AGYEJ
AHHHB
AHJVU
AIEXJ
AIKHN
AITUG
AJBFU
AJOXV
ALMA_UNASSIGNED_HOLDINGS
AMFUW
AMRAJ
AXJTR
AZFZN
BKOJK
BLXMC
CS3
EBS
EFJIC
EJD
EO8
EO9
EP2
EP3
FDB
FGOYB
FIRID
FNPLU
FYGXN
G-Q
G8K
GBOLZ
HLZ
HZ~
IHE
JJJVA
KOM
LG9
M41
MO0
O-L
O9-
OAUVE
P-9
PC.
Q38
R2-
RIG
ROL
RPZ
SBC
SDF
SPC
SST
SSV
SSZ
T5K
TN5
UHS
WUQ
YYP
AATTM
AAXKI
AAYWO
AAYXX
ABDPE
ABJNI
ABWVN
ACLOT
ACRPL
ACVFH
ADCNI
ADNMO
AEIPS
AEUPX
AFPUW
AGQPQ
AIGII
AIIUN
AKBMS
AKRWK
AKYEP
ANKPU
CITATION
EFKBS
~HD
7SC
8FD
JQ2
L7M
L~C
L~D
ID FETCH-LOGICAL-c335t-348435f302faea582c1611493b7d22cb4abb2d509b83f1036dd7cf44bd507cc73
IEDL.DBID AIKHN
ISSN 0165-6074
IngestDate Sat Sep 27 17:03:23 EDT 2025
Thu Apr 24 22:56:31 EDT 2025
Wed Oct 01 02:37:33 EDT 2025
Fri Feb 23 02:28:51 EST 2024
IsPeerReviewed false
IsScholarly false
Issue 10
Keywords Neuroprocessors
Fixed-point format
Neural networks
Language English
License https://www.elsevier.com/tdm/userlicense/1.0
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c335t-348435f302faea582c1611493b7d22cb4abb2d509b83f1036dd7cf44bd507cc73
Notes ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Feature-1
content type line 23
PQID 26209666
PQPubID 23500
PageCount 13
ParticipantIDs proquest_miscellaneous_26209666
crossref_citationtrail_10_1016_0165_6074_96_00012_9
crossref_primary_10_1016_0165_6074_96_00012_9
elsevier_sciencedirect_doi_10_1016_0165_6074_96_00012_9
ProviderPackageCode CITATION
AAYXX
PublicationCentury 1900
PublicationDate 1996-06-01
PublicationDateYYYYMMDD 1996-06-01
PublicationDate_xml – month: 06
  year: 1996
  text: 1996-06-01
  day: 01
PublicationDecade 1990
PublicationTitle Microprocessing and microprogramming
PublicationYear 1996
Publisher Elsevier B.V
Publisher_xml – name: Elsevier B.V
References Ramacher (BIB21) Oct. 1991
Bourlard, Morgan (BIB7) Nov. 1993; 4
Caviglia, Valle, Bisio (BIB10) 17–21 June 1990
Asanović, Beck, Feldman, Morgan, Wawrzynek (BIB4) Dec. 1993; 4
(BIB20) 1991
Bochev (BIB6) May 1993; 41
Hammerstrom (BIB13) 17–21 June 1990
Sirat, Makram-Ebeid, Zorer, Nadal (BIB23) 1989
Asanović, Beck, Irissou, Kingsbury, Morgan, Wawrzynek (BIB3) 13–15 Aug. 1995
Cybenko (BIB9) 1989; 2
Kane, Heinrich (BIB17) 1992
Sakaue, Kohda, Yamamoto, Maruno, Shimeki (BIB22) March 1993; 4
Sejnowsky, Rosenberg (BIB24) 1987; 1
E. Fiesler, A. Choudry and H.J. Caulfield, A universal weight discretization method for multi-layer neural networks
to appear.
Tollenaere (BIB26) 1990; 3
Vogl, Mangis, Rigler, Zink, Alkon (BIB27) 1989; 59
Hollis, Harper, Paulos (BIB15) 1990; 2
Asanović, Morgan (BIB5) 16–18 Oct. 1991
Maudit, Duranton, Gobert, Sirat (BIB18) May 1992; 3
Kramer (BIB16) Feb. 1991; 37
Anguita, Parodi, Zunino (BIB2) 1994; 6
Carrato, Premoli, Sicuranza (BIB8) 1991
Tesauro, Sejnowsky (BIB25) 1987
Hoehfeld, Fahlman (BIB14) July 1992; 3
Morgan, Bourlard (BIB19) Nov. 1992; 4
Wawrzynek, Asanović, Morgan (BIB28) May 1993; 4
Alippi, Negri (BIB1) 1991
Corana, Rolando, Ridella (BIB11) 1989
Bochev (10.1016/0165-6074(96)00012-9_BIB6) 1993; 41
Asanović (10.1016/0165-6074(96)00012-9_BIB3) 1995
Tesauro (10.1016/0165-6074(96)00012-9_BIB25) 1987
Caviglia (10.1016/0165-6074(96)00012-9_BIB10) 1990
Asanović (10.1016/0165-6074(96)00012-9_BIB4) 1993; 4
Sirat (10.1016/0165-6074(96)00012-9_BIB23) 1989
Alippi (10.1016/0165-6074(96)00012-9_BIB1) 1991
(10.1016/0165-6074(96)00012-9_BIB20) 1991
Sejnowsky (10.1016/0165-6074(96)00012-9_BIB24) 1987; 1
Corana (10.1016/0165-6074(96)00012-9_BIB11) 1989
Wawrzynek (10.1016/0165-6074(96)00012-9_BIB28) 1993; 4
Tollenaere (10.1016/0165-6074(96)00012-9_BIB26) 1990; 3
10.1016/0165-6074(96)00012-9_BIB12
Maudit (10.1016/0165-6074(96)00012-9_BIB18) 1992; 3
Hollis (10.1016/0165-6074(96)00012-9_BIB15) 1990; 2
Kramer (10.1016/0165-6074(96)00012-9_BIB16) 1991; 37
Hoehfeld (10.1016/0165-6074(96)00012-9_BIB14) 1992; 3
Vogl (10.1016/0165-6074(96)00012-9_BIB27) 1989; 59
Carrato (10.1016/0165-6074(96)00012-9_BIB8) 1991
Kane (10.1016/0165-6074(96)00012-9_BIB17) 1992
Ramacher (10.1016/0165-6074(96)00012-9_BIB21) 1991
Asanović (10.1016/0165-6074(96)00012-9_BIB5) 1991
Bourlard (10.1016/0165-6074(96)00012-9_BIB7) 1993; 4
Cybenko (10.1016/0165-6074(96)00012-9_BIB9) 1989; 2
Anguita (10.1016/0165-6074(96)00012-9_BIB2) 1994; 6
Sakaue (10.1016/0165-6074(96)00012-9_BIB22) 1993; 4
Hammerstrom (10.1016/0165-6074(96)00012-9_BIB13) 1990
Morgan (10.1016/0165-6074(96)00012-9_BIB19) 1992; 4
References_xml – volume: 4
  start-page: 893
  year: Nov. 1993
  end-page: 909
  ident: BIB7
  article-title: Continuous speech recognition by connectionist statistical methods
  publication-title: IEEE Trans. on Neural Networks
– volume: 4
  start-page: 394
  year: May 1993
  end-page: 399
  ident: BIB28
  article-title: The design of a neuro-microprocessor
  publication-title: IEEE Trans. on Neural Networks
– volume: 3
  start-page: 414
  year: May 1992
  end-page: 421
  ident: BIB18
  article-title: Lneuro 1.0: a piece of hardware LEGO for building neural network systems
  publication-title: IEEE Trans. on Neural Networks
– volume: 59
  start-page: 257
  year: 1989
  end-page: 263
  ident: BIB27
  article-title: Accelerating the covergence of the back-propagation method
  publication-title: Biological Cybernetics
– volume: 2
  year: 1990
  ident: BIB15
  article-title: The effect of precision constraints in a backpropagation learning network
  publication-title: Neural Computation
– year: 1992
  ident: BIB17
  publication-title: MIPS RISC Architecture
– start-page: 631
  year: 17–21 June 1990
  end-page: 637
  ident: BIB10
  article-title: Effect of weight discretization on the back propagation learning method: Algorithm design and hardware realization
  publication-title: Proc. of IJCNN '90
– start-page: 537
  year: 17–21 June 1990
  end-page: 544
  ident: BIB13
  article-title: A VLSI architecture for high-performance, low-cost, on-chip learning
  publication-title: Proc. of the IJCNN '90
– volume: 4
  start-page: 317
  year: Dec. 1993
  end-page: 326
  ident: BIB4
  article-title: Designing a connectionist network supercomputer
  publication-title: Int. J. Neural Systems
– volume: 6
  start-page: 57
  year: 1994
  end-page: 65
  ident: BIB2
  article-title: An efficient implementation of BP on RISC-based workstations
  publication-title: Neurocomputing
– volume: 4
  start-page: 270
  year: March 1993
  end-page: 275
  ident: BIB22
  article-title: Reduction of required precision bits for back-propagation applied to pattern recognition
  publication-title: IEEE Trans. on Neural Networks
– volume: 4
  start-page: 835
  year: Nov. 1992
  end-page: 838
  ident: BIB19
  article-title: Factoring networks by a statistical method
  publication-title: Neural Computation
– start-page: 442
  year: 1987
  end-page: 456
  ident: BIB25
  article-title: A neural network that learns to play backgammon
  publication-title: Neural Information Processing Systems
– start-page: 181
  year: 1989
  end-page: 185
  ident: BIB23
  article-title: Unlimited accuracy in layered networks
  publication-title: IEEE Int. Conf. on Artificial Neural Networks
– volume: 1
  start-page: 145
  year: 1987
  end-page: 168
  ident: BIB24
  article-title: Parallel networks that learn to pronounce English text
  publication-title: Complex Systems
– start-page: 9
  year: 16–18 Oct. 1991
  end-page: 15
  ident: BIB5
  article-title: Experimental determination of precision requirements for back-propagation training of artificial neural networks
  publication-title: Proc. of 2nd Int. Conf. on Microelectronics for Neural Networks
– start-page: 526
  year: 1991
  end-page: 531
  ident: BIB8
  article-title: Linear and nonlinear neural networks for image compression
  publication-title: Digital Signal Processing
– volume: 3
  start-page: 561
  year: 1990
  end-page: 573
  ident: BIB26
  article-title: SuperSAB: fast adaptive back propagation with good scaling properties
  publication-title: Neural Networks
– volume: 41
  year: May 1993
  ident: BIB6
  article-title: Distributed arithmetic implementation of artificial neural networks
  publication-title: IEEE Trans. on Signal Processing
– volume: 37
  start-page: 233
  year: Feb. 1991
  end-page: 243
  ident: BIB16
  article-title: Nonlinear principal component analysis using autoassociative neural networks
  publication-title: AIChE J.
– volume: 3
  start-page: 602
  year: July 1992
  end-page: 611
  ident: BIB14
  article-title: Learning with numerical precision using the cascade-correlation algorithm
  publication-title: IEEE Trans. on Neural Networks
– volume: 2
  start-page: 303
  year: 1989
  end-page: 314
  ident: BIB9
  article-title: Approximation by superposition of a sigmoidal function
  publication-title: Math of Control, Signal, and Systems
– year: 13–15 Aug. 1995
  ident: BIB3
  article-title: The TO vector microprocessor
  publication-title: Hot Chips VII Symposium
– reference: , to appear.
– reference: E. Fiesler, A. Choudry and H.J. Caulfield, A universal weight discretization method for multi-layer neural networks,
– start-page: 181
  year: 1989
  end-page: 190
  ident: BIB11
  article-title: A highly efficient implementation of back-propagation algorithm on SIMD computers
  publication-title: High Performance Computing, Proc. of the Int. Symp.
– year: 1991
  ident: BIB20
  publication-title: VLSI Design of Neural Networks
– start-page: 1873
  year: 1991
  end-page: 1878
  ident: BIB1
  article-title: Hardware requirements for digital VLSI implementations of neural networks
  publication-title: Int. Joint Conf. on Neural Networks
– start-page: 401
  year: Oct. 1991
  end-page: 409
  ident: BIB21
  article-title: SYNAPSE-X: a general-purpose neurocomputer
  publication-title: Proc. of the 2nd Int. Conf. on Microelectronics for Neural Networks
– start-page: 526
  year: 1991
  ident: 10.1016/0165-6074(96)00012-9_BIB8
  article-title: Linear and nonlinear neural networks for image compression
– volume: 1
  start-page: 145
  year: 1987
  ident: 10.1016/0165-6074(96)00012-9_BIB24
  article-title: Parallel networks that learn to pronounce English text
  publication-title: Complex Systems
– volume: 3
  start-page: 602
  issue: 4
  year: 1992
  ident: 10.1016/0165-6074(96)00012-9_BIB14
  article-title: Learning with numerical precision using the cascade-correlation algorithm
  publication-title: IEEE Trans. on Neural Networks
  doi: 10.1109/72.143374
– year: 1995
  ident: 10.1016/0165-6074(96)00012-9_BIB3
  article-title: The TO vector microprocessor
– start-page: 631
  year: 1990
  ident: 10.1016/0165-6074(96)00012-9_BIB10
  article-title: Effect of weight discretization on the back propagation learning method: Algorithm design and hardware realization
– volume: 59
  start-page: 257
  year: 1989
  ident: 10.1016/0165-6074(96)00012-9_BIB27
  article-title: Accelerating the covergence of the back-propagation method
  publication-title: Biological Cybernetics
  doi: 10.1007/BF00332914
– start-page: 442
  year: 1987
  ident: 10.1016/0165-6074(96)00012-9_BIB25
  article-title: A neural network that learns to play backgammon
– start-page: 181
  year: 1989
  ident: 10.1016/0165-6074(96)00012-9_BIB23
  article-title: Unlimited accuracy in layered networks
– volume: 41
  issue: 5
  year: 1993
  ident: 10.1016/0165-6074(96)00012-9_BIB6
  article-title: Distributed arithmetic implementation of artificial neural networks
  publication-title: IEEE Trans. on Signal Processing
  doi: 10.1109/78.215327
– start-page: 181
  year: 1989
  ident: 10.1016/0165-6074(96)00012-9_BIB11
  article-title: A highly efficient implementation of back-propagation algorithm on SIMD computers
– volume: 3
  start-page: 561
  issue: 5
  year: 1990
  ident: 10.1016/0165-6074(96)00012-9_BIB26
  article-title: SuperSAB: fast adaptive back propagation with good scaling properties
  publication-title: Neural Networks
  doi: 10.1016/0893-6080(90)90006-7
– volume: 4
  start-page: 835
  issue: 6
  year: 1992
  ident: 10.1016/0165-6074(96)00012-9_BIB19
  article-title: Factoring networks by a statistical method
  publication-title: Neural Computation
  doi: 10.1162/neco.1992.4.6.835
– volume: 3
  start-page: 414
  issue: 3
  year: 1992
  ident: 10.1016/0165-6074(96)00012-9_BIB18
  article-title: Lneuro 1.0: a piece of hardware LEGO for building neural network systems
  publication-title: IEEE Trans. on Neural Networks
  doi: 10.1109/72.129414
– year: 1991
  ident: 10.1016/0165-6074(96)00012-9_BIB20
– start-page: 1873
  year: 1991
  ident: 10.1016/0165-6074(96)00012-9_BIB1
  article-title: Hardware requirements for digital VLSI implementations of neural networks
– volume: 2
  issue: 3
  year: 1990
  ident: 10.1016/0165-6074(96)00012-9_BIB15
  article-title: The effect of precision constraints in a backpropagation learning network
  publication-title: Neural Computation
  doi: 10.1162/neco.1990.2.3.363
– start-page: 9
  year: 1991
  ident: 10.1016/0165-6074(96)00012-9_BIB5
  article-title: Experimental determination of precision requirements for back-propagation training of artificial neural networks
– volume: 2
  start-page: 303
  year: 1989
  ident: 10.1016/0165-6074(96)00012-9_BIB9
  article-title: Approximation by superposition of a sigmoidal function
  publication-title: Math of Control, Signal, and Systems
  doi: 10.1007/BF02551274
– year: 1992
  ident: 10.1016/0165-6074(96)00012-9_BIB17
– start-page: 537
  year: 1990
  ident: 10.1016/0165-6074(96)00012-9_BIB13
  article-title: A VLSI architecture for high-performance, low-cost, on-chip learning
– volume: 37
  start-page: 233
  issue: 2
  year: 1991
  ident: 10.1016/0165-6074(96)00012-9_BIB16
  article-title: Nonlinear principal component analysis using autoassociative neural networks
  publication-title: AIChE J.
  doi: 10.1002/aic.690370209
– start-page: 401
  year: 1991
  ident: 10.1016/0165-6074(96)00012-9_BIB21
  article-title: SYNAPSE-X: a general-purpose neurocomputer
– volume: 4
  start-page: 317
  issue: 4
  year: 1993
  ident: 10.1016/0165-6074(96)00012-9_BIB4
  article-title: Designing a connectionist network supercomputer
  publication-title: Int. J. Neural Systems
  doi: 10.1142/S0129065793000250
– volume: 4
  start-page: 270
  issue: 2
  year: 1993
  ident: 10.1016/0165-6074(96)00012-9_BIB22
  article-title: Reduction of required precision bits for back-propagation applied to pattern recognition
  publication-title: IEEE Trans. on Neural Networks
  doi: 10.1109/72.207614
– volume: 6
  start-page: 57
  year: 1994
  ident: 10.1016/0165-6074(96)00012-9_BIB2
  article-title: An efficient implementation of BP on RISC-based workstations
  publication-title: Neurocomputing
  doi: 10.1016/0925-2312(94)90034-5
– ident: 10.1016/0165-6074(96)00012-9_BIB12
– volume: 4
  start-page: 394
  issue: 3
  year: 1993
  ident: 10.1016/0165-6074(96)00012-9_BIB28
  article-title: The design of a neuro-microprocessor
  publication-title: IEEE Trans. on Neural Networks
  doi: 10.1109/72.217180
– volume: 4
  start-page: 893
  issue: 6
  year: 1993
  ident: 10.1016/0165-6074(96)00012-9_BIB7
  article-title: Continuous speech recognition by connectionist statistical methods
  publication-title: IEEE Trans. on Neural Networks
  doi: 10.1109/72.286885
SSID ssj0004975
Score 1.2609348
Snippet We examine the efficient implementation of back-propagation (BP) type algorithms on TO [3], a vector processor with a fixed-point engine, designed for neural...
We examine the efficient implementation of back-propagation (BP) type algorithms on T0 [3], a vector processor with a fixed-point engine, designed for neural...
SourceID proquest
crossref
elsevier
SourceType Aggregation Database
Enrichment Source
Index Database
Publisher
StartPage 757
SubjectTerms Fixed-point format
Neural networks
Neuroprocessors
Title Mixing floating- and fixed-point formats for neural network learning on neuroprocessors
URI https://dx.doi.org/10.1016/0165-6074(96)00012-9
https://www.proquest.com/docview/26209666
Volume 41
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVESC
  databaseName: Elsevier SD Complete Freedom Collection
  issn: 0165-6074
  databaseCode: ACRLP
  dateStart: 19950401
  customDbUrl:
  isFulltext: true
  dateEnd: 19960630
  titleUrlDefault: https://www.sciencedirect.com
  omitProxy: true
  ssIdentifier: ssj0004975
  providerName: Elsevier
– providerCode: PRVESC
  databaseName: Elsevier SD Freedom Collection
  issn: 0165-6074
  databaseCode: .~1
  dateStart: 19950101
  customDbUrl:
  isFulltext: true
  dateEnd: 19960630
  titleUrlDefault: https://www.sciencedirect.com
  omitProxy: true
  ssIdentifier: ssj0004975
  providerName: Elsevier
– providerCode: PRVESC
  databaseName: Elsevier SD Freedom Collection Journals [SCFCJ]
  issn: 0165-6074
  databaseCode: AIKHN
  dateStart: 19950401
  customDbUrl:
  isFulltext: true
  dateEnd: 19960630
  titleUrlDefault: https://www.sciencedirect.com
  omitProxy: true
  ssIdentifier: ssj0004975
  providerName: Elsevier
– providerCode: PRVLSH
  databaseName: Elsevier Journals
  issn: 0165-6074
  databaseCode: AKRWK
  dateStart: 19950401
  customDbUrl:
  isFulltext: true
  mediaType: online
  dateEnd: 99991231
  omitProxy: true
  ssIdentifier: ssj0004975
  providerName: Library Specific Holdings
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV1LT9wwELbY5dJLebRVKbD4wKE9uJvEjp0cEQItT1Woq-7Nsh27ioSSFbtInPjtjGOHAhJaqadIk4wVjT0zn-15IHRYZZnNGU8IQPGEMCdyohinhKW6EDZLtVV-o3h1zSdTdj7LZy9yYXxYZbT9waZ31jpSxlGa43ldj5OuboxgAMG705QBWgf3UxRDtH50djG5_pcdWYoQyMhz4hn6BLqUj59p30v-oxuGlO85qDemuvM_p5voYwSO-Cj82xZas8022uibMuCoo5_Qn6v6AdwRdret8iHNBKumwq5-sBWZt3WzxAGmLvwT-3KWMGoTgsFx7CHxF7dN96qdhzyC9m7xGU1PT34fT0jsnkAMpfmSUFYAFHI0yZyyKi8yA-AO9kNUC5ggo5nSOqsAL-iCuhQcWVUJ4xjTQBPGCPoFDZu2sV8RdoLZxChQ1dQywbVOSpumVWkTR3lh9A6ivcikiaXFfYeLW9nHkHlBSy9oWfJw3y3LHUSeueahtMaK70U_G_LVEpFg_VdwHvSTJ0F7_JWIamx7v5C-HD9s-Pi3_x57F30Igdz-YGYPDZd393YfcMpSj9Dg52M6guV4fHP5axSX5RNb6-L1
linkProvider Elsevier
linkToHtml http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV1LT8QgECa6HvTi2_iWgwc94LaFQns0RrM-dk8avRGgYJps2ma3JnvytwulXR-JMfHUhAIhA8x8Ax8zAJxmUaRjQgNkoXiAiGExEoRiREKZMB2FUgvnKA5HdPBE7l7ily9vYRytstX9Xqc32rot6bfS7Fd53g-auDGMWAjenKYsgiUSR8w5YBfvnzQPkjJPY6QxctW753Mh7c_LzlJ63nSC0t_M0w9F3Vifm3Ww2sJGeOlHtgEWdLEJ1rqUDLDdoVvgeZjPrDGCZlwKR2hGUBQZNPlMZ6gq86KGHqRO3Re6YJa218JTwWGbQeIVlkXzq6z8K4JyMt0GTzfXj1cD1OZOQArjuEaYJBYIGRxERmgRJ5Gy0M56Q1gyOz1KEiFllFm0IBNsQmvGsowpQ4i0ZUwphndArygLvQugYUQHStiNGmrCqJRBqsMwS3VgME2U3AO4ExlXbWBxl99izDsGmRM0d4LmKfW33TzdA2jeqvKBNf6oz7rZ4N8WCLe6_4-WJ93kcbt33IWIKHT5NuUuGL919-j-v_s-AcuDx-EDf7gd3R-AFU_pdkc0h6BXT970kUUstTxuluQHHxniKA
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=Mixing+floating-+and+fixed-point+formats+for+neural+network+learning+on+neuroprocessors&rft.jtitle=Microprocessing+and+microprogramming&rft.au=Anguita%2C+Davide&rft.au=Gomes%2C+Benedict+A.&rft.date=1996-06-01&rft.pub=Elsevier+B.V&rft.issn=0165-6074&rft.volume=41&rft.issue=10&rft.spage=757&rft.epage=769&rft_id=info:doi/10.1016%2F0165-6074%2896%2900012-9&rft.externalDocID=0165607496000129
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0165-6074&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0165-6074&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0165-6074&client=summon