Time series generation by recurrent neural networks

A. Priel, I. Kanter

Research output: Contribution to journalArticlepeer-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 languageEnglish
Pages (from-to)315-332
Number of pages18
JournalAnnals of Mathematics and Artificial Intelligence
Volume39
Issue number3
DOIs
StatePublished - Nov 2003

Bibliographical note

Funding Information:
IK acknowledges partial support from the GIF.

Funding

IK acknowledges partial support from the GIF.

FundersFunder 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

    Fingerprint

    Dive into the research topics of 'Time series generation by recurrent neural networks'. Together they form a unique fingerprint.

    Cite this