The learning dynamics of a universal approximator

Ansgar H. L. West; David Saad; Ian T. Nabney; Michael C. Mozer; Thomas Petsche; Michael I. Jordan

The learning dynamics of a universal approximator

Ansgar H. L. West, David Saad, Ian T. Nabney, Michael C. Mozer (Editor), Thomas Petsche (Editor), Michael I. Jordan (Editor)

Computer Science Research Group

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

Abstract

The learning properties of a universal approximator, a normalized committee machine with adjustable biases, are studied for on-line back-propagation learning. Within a statistical mechanics framework, numerical studies show that this model has features which do not exist in previously studied two-layer network models without adjustable biases, e.g., attractive suboptimal symmetric phases even for realizable cases and noiseless data.

Original language	English
Pages (from-to)	288-294
Number of pages	7
Journal	Advances in Neural Information Processing Systems
Volume	9
Publication status	Published - May 1997

Bibliographical note

Copyright of the Massachusetts Institute of Technology Press (MIT Press)

Keywords

approximator
back-propagation
symmetric phases
realizable cases
noiseless data

Access to Document

The learning dynamics of a universal approximator
Copyright of the Massachusetts Institute of Technology Press (MIT Press)
Final published version, 786 KB

Cite this

@article{daa79ae7900145ad9d24505409375016,

title = "The learning dynamics of a universal approximator",

abstract = "The learning properties of a universal approximator, a normalized committee machine with adjustable biases, are studied for on-line back-propagation learning. Within a statistical mechanics framework, numerical studies show that this model has features which do not exist in previously studied two-layer network models without adjustable biases, e.g., attractive suboptimal symmetric phases even for realizable cases and noiseless data.",

keywords = "approximator, back-propagation, symmetric phases, realizable cases, noiseless data",

author = "West, {Ansgar H. L.} and David Saad and Nabney, {Ian T.} and Mozer, {Michael C.} and Thomas Petsche and Jordan, {Michael I.}",

note = "Copyright of the Massachusetts Institute of Technology Press (MIT Press)",

year = "1997",

month = may,

language = "English",

volume = "9",

pages = "288--294",

journal = "Advances in Neural Information Processing Systems",

issn = "1049-5258",

publisher = "MIT Press Journals",

}

TY - JOUR

T1 - The learning dynamics of a universal approximator

AU - West, Ansgar H. L.

AU - Saad, David

AU - Nabney, Ian T.

A2 - Mozer, Michael C.

A2 - Petsche, Thomas

A2 - Jordan, Michael I.

N1 - Copyright of the Massachusetts Institute of Technology Press (MIT Press)

PY - 1997/5

Y1 - 1997/5

N2 - The learning properties of a universal approximator, a normalized committee machine with adjustable biases, are studied for on-line back-propagation learning. Within a statistical mechanics framework, numerical studies show that this model has features which do not exist in previously studied two-layer network models without adjustable biases, e.g., attractive suboptimal symmetric phases even for realizable cases and noiseless data.

AB - The learning properties of a universal approximator, a normalized committee machine with adjustable biases, are studied for on-line back-propagation learning. Within a statistical mechanics framework, numerical studies show that this model has features which do not exist in previously studied two-layer network models without adjustable biases, e.g., attractive suboptimal symmetric phases even for realizable cases and noiseless data.

KW - approximator

KW - back-propagation

KW - symmetric phases

KW - realizable cases

KW - noiseless data

UR - http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=3990

UR - http://www.scopus.com/inward/record.url?scp=84899020045&partnerID=8YFLogxK

M3 - Article

SN - 1049-5258

VL - 9

SP - 288

EP - 294

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

ER -

The learning dynamics of a universal approximator

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this