Multi-scale training of a large backpropagation net

E. W. Feiten; O. Martin; S. W. Otto; J. Hutchinson

doi:10.1007/BF00205112

Multi-scale training of a large backpropagation net

E. W. Feiten
, O. Martin
, S. W. Otto
, J. Hutchinson

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

The shortcomings of automatic learning in neural networks can be overcome by incorporating features of the problem into the weight space. We teach a multi-layer perceptron to recognize hand-printed characters by using a multi-scale method. This leads to better performance for learning rates and for generalization than direct use of back-propagation.

Original language	English (US)
Pages (from-to)	503-509
Number of pages	7
Journal	Biological Cybernetics
Volume	62
Issue number	6
DOIs	https://doi.org/10.1007/BF00205112
State	Published - Apr 1990

All Science Journal Classification (ASJC) codes

Biotechnology
General Computer Science

Access to Document

10.1007/BF00205112

Cite this

@article{ee77b4d9480e4d8f84f1ccb8395b7e95,

title = "Multi-scale training of a large backpropagation net",

abstract = "The shortcomings of automatic learning in neural networks can be overcome by incorporating features of the problem into the weight space. We teach a multi-layer perceptron to recognize hand-printed characters by using a multi-scale method. This leads to better performance for learning rates and for generalization than direct use of back-propagation.",

author = "Feiten, \{E. W.\} and O. Martin and Otto, \{S. W.\} and J. Hutchinson",

year = "1990",

month = apr,

doi = "10.1007/BF00205112",

language = "English (US)",

volume = "62",

pages = "503--509",

journal = "Biological Cybernetics",

issn = "0340-1200",

publisher = "Springer Science and Business Media Deutschland GmbH",

number = "6",

}

TY - JOUR

T1 - Multi-scale training of a large backpropagation net

AU - Feiten, E. W.

AU - Martin, O.

AU - Otto, S. W.

AU - Hutchinson, J.

PY - 1990/4

Y1 - 1990/4

N2 - The shortcomings of automatic learning in neural networks can be overcome by incorporating features of the problem into the weight space. We teach a multi-layer perceptron to recognize hand-printed characters by using a multi-scale method. This leads to better performance for learning rates and for generalization than direct use of back-propagation.

AB - The shortcomings of automatic learning in neural networks can be overcome by incorporating features of the problem into the weight space. We teach a multi-layer perceptron to recognize hand-printed characters by using a multi-scale method. This leads to better performance for learning rates and for generalization than direct use of back-propagation.

UR - https://www.scopus.com/pages/publications/0025299836

UR - https://www.scopus.com/pages/publications/0025299836#tab=citedBy

U2 - 10.1007/BF00205112

DO - 10.1007/BF00205112

M3 - Article

AN - SCOPUS:0025299836

SN - 0340-1200

VL - 62

SP - 503

EP - 509

JO - Biological Cybernetics

JF - Biological Cybernetics

IS - 6

ER -

Multi-scale training of a large backpropagation net

Abstract

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this