Dynamics of learning with restricted training sets

Anthony C.C. Coolen; David Saad

doi:10.1103/PhysRevE.62.5444

Dynamics of learning with restricted training sets

Anthony C.C. Coolen, David Saad

Aston University

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

Abstract

The dynamics of supervised learning in layered neural networks were studied in the regime where the size of the training set is proportional to the number of inputs. The evolution of macroscopic observables, including the two relevant performance measures can be predicted by using the dynamical replica theory. Three approximation schemes aimed at eliminating the need to solve a functional saddle-point equation at each time step have been derived.

Original language	English
Pages (from-to)	5444-5487
Number of pages	44
Journal	Physical Review E
Volume	62
Issue number	4
DOIs	https://doi.org/10.1103/PhysRevE.62.5444
Publication status	Published - Oct 2000

Bibliographical note

Copyright of American Physical Society

Keywords

layered neural networks
learning dynamics
dynamically replica theory

Access to Document

10.1103/PhysRevE.62.5444

NCRG_2000_005.pdf
Copyright of American Physical Society
Accepted author manuscript, 12.7 MB

Cite this

@article{fcefd2b6f87a4bd5a777d08cbb351678,

title = "Dynamics of learning with restricted training sets",

abstract = "The dynamics of supervised learning in layered neural networks were studied in the regime where the size of the training set is proportional to the number of inputs. The evolution of macroscopic observables, including the two relevant performance measures can be predicted by using the dynamical replica theory. Three approximation schemes aimed at eliminating the need to solve a functional saddle-point equation at each time step have been derived.",

keywords = "layered neural networks, learning dynamics, dynamically replica theory",

author = "Coolen, {Anthony C.C.} and David Saad",

note = "Copyright of American Physical Society",

year = "2000",

month = oct,

doi = "10.1103/PhysRevE.62.5444",

language = "English",

volume = "62",

pages = "5444--5487",

journal = "Physical Review E",

issn = "1063-651X",

publisher = "American Physical Society",

number = "4",

}

TY - JOUR

T1 - Dynamics of learning with restricted training sets

AU - Coolen, Anthony C.C.

AU - Saad, David

N1 - Copyright of American Physical Society

PY - 2000/10

Y1 - 2000/10

N2 - The dynamics of supervised learning in layered neural networks were studied in the regime where the size of the training set is proportional to the number of inputs. The evolution of macroscopic observables, including the two relevant performance measures can be predicted by using the dynamical replica theory. Three approximation schemes aimed at eliminating the need to solve a functional saddle-point equation at each time step have been derived.

AB - The dynamics of supervised learning in layered neural networks were studied in the regime where the size of the training set is proportional to the number of inputs. The evolution of macroscopic observables, including the two relevant performance measures can be predicted by using the dynamical replica theory. Three approximation schemes aimed at eliminating the need to solve a functional saddle-point equation at each time step have been derived.

KW - layered neural networks

KW - learning dynamics

KW - dynamically replica theory

UR - http://www.scopus.com/inward/record.url?scp=0034298111&partnerID=8YFLogxK

UR - http://link.aps.org/doi/10.1103/PhysRevE.62.5444

U2 - 10.1103/PhysRevE.62.5444

DO - 10.1103/PhysRevE.62.5444

M3 - Article

SN - 1063-651X

VL - 62

SP - 5444

EP - 5487

JO - Physical Review E

JF - Physical Review E

IS - 4

ER -

Dynamics of learning with restricted training sets

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this