Back-propagation is significantly faster if the expected value of the source unit is used for update

Tariq Samad

doi:10.1016/0893-6080(88)90253-5

Back-propagation is significantly faster if the expected value of the source unit is used for update

Tariq Samad

Research output: Contribution to journal › Conference article › peer-review

8 Scopus citations

Abstract

It has frequently been remarked that a major drawback of the back-propagation learning rule, and one that does not bode well for its application to real-world problems, is its poor scaling properties - with large networks, back-propagation can take infeasibly long to converge. The research outlined shows that an intuitively straightforward modification of back-propagation can greatly improve its performance, particularly for large and structured networks.

Original language	English (US)
Pages (from-to)	216
Number of pages	1
Journal	Neural Networks
Volume	1
Issue number	1 SUPPL
DOIs	https://doi.org/10.1016/0893-6080(88)90253-5
State	Published - 1988
Externally published	Yes
Event	International Neural Network Society 1988 First Annual Meeting - Boston, MA, USA Duration: Sep 6 1988 → Sep 10 1988

Publisher link

10.1016/0893-6080(88)90253-5

Find It

Find It at U of M Libraries

Cite this

@article{d902555327d64695989c7f8bf88cc608,

title = "Back-propagation is significantly faster if the expected value of the source unit is used for update",

abstract = "It has frequently been remarked that a major drawback of the back-propagation learning rule, and one that does not bode well for its application to real-world problems, is its poor scaling properties - with large networks, back-propagation can take infeasibly long to converge. The research outlined shows that an intuitively straightforward modification of back-propagation can greatly improve its performance, particularly for large and structured networks.",

author = "Tariq Samad",

year = "1988",

doi = "10.1016/0893-6080(88)90253-5",

language = "English (US)",

volume = "1",

pages = "216",

journal = "Neural Networks",

issn = "0893-6080",

publisher = "Elsevier Ltd",

number = "1 SUPPL",

note = "International Neural Network Society 1988 First Annual Meeting ; Conference date: 06-09-1988 Through 10-09-1988",

}

TY - JOUR

T1 - Back-propagation is significantly faster if the expected value of the source unit is used for update

AU - Samad, Tariq

PY - 1988

Y1 - 1988

N2 - It has frequently been remarked that a major drawback of the back-propagation learning rule, and one that does not bode well for its application to real-world problems, is its poor scaling properties - with large networks, back-propagation can take infeasibly long to converge. The research outlined shows that an intuitively straightforward modification of back-propagation can greatly improve its performance, particularly for large and structured networks.

AB - It has frequently been remarked that a major drawback of the back-propagation learning rule, and one that does not bode well for its application to real-world problems, is its poor scaling properties - with large networks, back-propagation can take infeasibly long to converge. The research outlined shows that an intuitively straightforward modification of back-propagation can greatly improve its performance, particularly for large and structured networks.

UR - https://www.scopus.com/pages/publications/0024167370

UR - https://www.scopus.com/pages/publications/0024167370#tab=citedBy

U2 - 10.1016/0893-6080(88)90253-5

DO - 10.1016/0893-6080(88)90253-5

M3 - Conference article

AN - SCOPUS:0024167370

SN - 0893-6080

VL - 1

SP - 216

JO - Neural Networks

JF - Neural Networks

IS - 1 SUPPL

T2 - International Neural Network Society 1988 First Annual Meeting

Y2 - 6 September 1988 through 10 September 1988

ER -

Back-propagation is significantly faster if the expected value of the source unit is used for update

Abstract

Publisher link

Find It

Other files and links

Fingerprint

Cite this