Time series generation by recurrent neural networks

A. Priel; I. Kanter

doi:10.1023/A:1024620813258

Time series generation by recurrent neural networks

A. Priel, I. Kanter

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

6 Scopus citations

Abstract

The properties of time series, generated by continuous valued feed-forward networks in which the next input vector is determined from past output values, are studied. Asymptotic solutions developed suggest that the typical stable behavior is (quasi) periodic with attractor dimension that is limited by the number of hidden units, independent of the details of the weights. The results are robust under additive noise, except for expected noise-induced effects - attractor broadening and loss of phase coherence at large times. These effects, however, are moderated by the size of the network N.

Original language	English
Pages (from-to)	315-332
Number of pages	18
Journal	Annals of Mathematics and Artificial Intelligence
Volume	39
Issue number	3
DOIs	https://doi.org/10.1023/A:1024620813258
State	Published - Nov 2003

Bibliographical note

Funding Information:
IK acknowledges partial support from the GIF.

Funding

IK acknowledges partial support from the GIF.

Funders	Funder number
German-Israeli Foundation for Scientific Research and Development

Keywords

Asymptotic properties
Attractor dimension
Non-linear dynamical systems
Recurrent neural networks
Stochastic processes
Time series

Access to Document

10.1023/A:1024620813258

Cite this

@article{ea7ca20818344faeaf136b3063546479,

title = "Time series generation by recurrent neural networks",

abstract = "The properties of time series, generated by continuous valued feed-forward networks in which the next input vector is determined from past output values, are studied. Asymptotic solutions developed suggest that the typical stable behavior is (quasi) periodic with attractor dimension that is limited by the number of hidden units, independent of the details of the weights. The results are robust under additive noise, except for expected noise-induced effects - attractor broadening and loss of phase coherence at large times. These effects, however, are moderated by the size of the network N.",

keywords = "Asymptotic properties, Attractor dimension, Non-linear dynamical systems, Recurrent neural networks, Stochastic processes, Time series",

author = "A. Priel and I. Kanter",

note = "Funding Information: IK acknowledges partial support from the GIF.",

year = "2003",

month = nov,

doi = "10.1023/A:1024620813258",

language = "אנגלית",

volume = "39",

pages = "315--332",

journal = "Annals of Mathematics and Artificial Intelligence",

issn = "1012-2443",

publisher = "Springer Netherlands",

number = "3",

}

TY - JOUR

T1 - Time series generation by recurrent neural networks

AU - Priel, A.

AU - Kanter, I.

N1 - Funding Information: IK acknowledges partial support from the GIF.

PY - 2003/11

Y1 - 2003/11

N2 - The properties of time series, generated by continuous valued feed-forward networks in which the next input vector is determined from past output values, are studied. Asymptotic solutions developed suggest that the typical stable behavior is (quasi) periodic with attractor dimension that is limited by the number of hidden units, independent of the details of the weights. The results are robust under additive noise, except for expected noise-induced effects - attractor broadening and loss of phase coherence at large times. These effects, however, are moderated by the size of the network N.

AB - The properties of time series, generated by continuous valued feed-forward networks in which the next input vector is determined from past output values, are studied. Asymptotic solutions developed suggest that the typical stable behavior is (quasi) periodic with attractor dimension that is limited by the number of hidden units, independent of the details of the weights. The results are robust under additive noise, except for expected noise-induced effects - attractor broadening and loss of phase coherence at large times. These effects, however, are moderated by the size of the network N.

KW - Asymptotic properties

KW - Attractor dimension

KW - Non-linear dynamical systems

KW - Recurrent neural networks

KW - Stochastic processes

KW - Time series

UR - http://www.scopus.com/inward/record.url?scp=0037829497&partnerID=8YFLogxK

U2 - 10.1023/A:1024620813258

DO - 10.1023/A:1024620813258

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???

AN - SCOPUS:0037829497

SN - 1012-2443

VL - 39

SP - 315

EP - 332

JO - Annals of Mathematics and Artificial Intelligence

JF - Annals of Mathematics and Artificial Intelligence

IS - 3

ER -

Time series generation by recurrent neural networks

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this