Critical echo state networks

Márton Albert Hajnal, A. Lőrincz

Research output: Chapter in Book/Report/Conference proceedingConference contribution

15 Citations (Scopus)

Abstract

We are interested in the optimization of the recurrent connection structure of Echo State Networks (ESNs), because their topology win strongly influence performance. We study ESN predictive capacity by numerical simulations 011 Mackey-Glass time series, and find that a particular small subset of ESNs is much better than ordinary ESNs provided that the topology of the recurrent feedback connections satisfies certain conditions. We argue that tire small subset separates two large sets of ESNs and this separation can be characterized in terms of phase transitions. With regard to the criticality of this phase transition, we introduce the notion of Critical Echo State Networks (CESN). We discuss why CESNs perform better than other ESNs.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages658-667
Number of pages10
Volume4131 LNCS - I
ISBN (Print)3540386254, 9783540386254
Publication statusPublished - 2006
Event16th International Conference on Artificial Neural Networks, ICANN 2006 - Athens, Greece
Duration: Sep 10 2006Sep 14 2006

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4131 LNCS - I
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other16th International Conference on Artificial Neural Networks, ICANN 2006
CountryGreece
CityAthens
Period9/10/069/14/06

Fingerprint

Echo State Network
Phase transitions
Topology
Tires
Time series
Feedback
Glass
Computer simulation
Phase Transition
Subset
Tire
Criticality
Large Set
Numerical Simulation
Optimization

Keywords

  • Critical point
  • Echo state network
  • Phase transition
  • Prediction
  • Time series

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Hajnal, M. A., & Lőrincz, A. (2006). Critical echo state networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4131 LNCS - I, pp. 658-667). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4131 LNCS - I). Springer Verlag.

Critical echo state networks. / Hajnal, Márton Albert; Lőrincz, A.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4131 LNCS - I Springer Verlag, 2006. p. 658-667 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4131 LNCS - I).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Hajnal, MA & Lőrincz, A 2006, Critical echo state networks. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 4131 LNCS - I, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 4131 LNCS - I, Springer Verlag, pp. 658-667, 16th International Conference on Artificial Neural Networks, ICANN 2006, Athens, Greece, 9/10/06.
Hajnal MA, Lőrincz A. Critical echo state networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4131 LNCS - I. Springer Verlag. 2006. p. 658-667. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Hajnal, Márton Albert ; Lőrincz, A. / Critical echo state networks. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 4131 LNCS - I Springer Verlag, 2006. pp. 658-667 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{8b883662555d44939c461c0225fcf2da,
title = "Critical echo state networks",
abstract = "We are interested in the optimization of the recurrent connection structure of Echo State Networks (ESNs), because their topology win strongly influence performance. We study ESN predictive capacity by numerical simulations 011 Mackey-Glass time series, and find that a particular small subset of ESNs is much better than ordinary ESNs provided that the topology of the recurrent feedback connections satisfies certain conditions. We argue that tire small subset separates two large sets of ESNs and this separation can be characterized in terms of phase transitions. With regard to the criticality of this phase transition, we introduce the notion of Critical Echo State Networks (CESN). We discuss why CESNs perform better than other ESNs.",
keywords = "Critical point, Echo state network, Phase transition, Prediction, Time series",
author = "Hajnal, {M{\'a}rton Albert} and A. Lőrincz",
year = "2006",
language = "English",
isbn = "3540386254",
volume = "4131 LNCS - I",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "658--667",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Critical echo state networks

AU - Hajnal, Márton Albert

AU - Lőrincz, A.

PY - 2006

Y1 - 2006

N2 - We are interested in the optimization of the recurrent connection structure of Echo State Networks (ESNs), because their topology win strongly influence performance. We study ESN predictive capacity by numerical simulations 011 Mackey-Glass time series, and find that a particular small subset of ESNs is much better than ordinary ESNs provided that the topology of the recurrent feedback connections satisfies certain conditions. We argue that tire small subset separates two large sets of ESNs and this separation can be characterized in terms of phase transitions. With regard to the criticality of this phase transition, we introduce the notion of Critical Echo State Networks (CESN). We discuss why CESNs perform better than other ESNs.

AB - We are interested in the optimization of the recurrent connection structure of Echo State Networks (ESNs), because their topology win strongly influence performance. We study ESN predictive capacity by numerical simulations 011 Mackey-Glass time series, and find that a particular small subset of ESNs is much better than ordinary ESNs provided that the topology of the recurrent feedback connections satisfies certain conditions. We argue that tire small subset separates two large sets of ESNs and this separation can be characterized in terms of phase transitions. With regard to the criticality of this phase transition, we introduce the notion of Critical Echo State Networks (CESN). We discuss why CESNs perform better than other ESNs.

KW - Critical point

KW - Echo state network

KW - Phase transition

KW - Prediction

KW - Time series

UR - http://www.scopus.com/inward/record.url?scp=33749836094&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33749836094&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:33749836094

SN - 3540386254

SN - 9783540386254

VL - 4131 LNCS - I

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 658

EP - 667

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -