Oscillatory associative memory network with perfect retrieval

Takashi Nishikawa, F C Hoppensteadt, Ying-Cheng Lai

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

Inspired by the discovery of possible roles of synchronization of oscillations in the brain, networks of coupled phase oscillators have been proposed before as models of associative memory based on the concept of temporal coding of information. Here we show, however, that error-free retrieval states of such networks turn out to be typically unstable regardless of the network size, in contrast to the classical Hopfield model. We propose a remedy for this undesirable property, and provide a systematic study of the improved model. In particular, we show that the error-free capacity of the network is at least 2epsilon(2) / log n patterns per neuron, where n is the number of oscillators (neurons) and epsilon the strength of the second-order mode in the coupling function. (C) 2004 Elsevier B.V. All rights reserved.

Original languageEnglish
Pages (from-to)134-148
Number of pages15
JournalPhysica. D, Nonlinear Phenomena
Volume197
Issue number1-2
DOIs
Publication statusPublished - 1 Oct 2004

Keywords

  • neural networks
  • phase oscillators
  • random matrices
  • model
  • natural frequencies
  • synchronization
  • dynamics
  • cortex

Cite this

Oscillatory associative memory network with perfect retrieval. / Nishikawa, Takashi; Hoppensteadt, F C ; Lai, Ying-Cheng.

In: Physica. D, Nonlinear Phenomena, Vol. 197, No. 1-2, 01.10.2004, p. 134-148.

Research output: Contribution to journalArticle

Nishikawa, Takashi ; Hoppensteadt, F C ; Lai, Ying-Cheng. / Oscillatory associative memory network with perfect retrieval. In: Physica. D, Nonlinear Phenomena. 2004 ; Vol. 197, No. 1-2. pp. 134-148.
@article{240e358d5dfe49d9997e2c7717afecbf,
title = "Oscillatory associative memory network with perfect retrieval",
abstract = "Inspired by the discovery of possible roles of synchronization of oscillations in the brain, networks of coupled phase oscillators have been proposed before as models of associative memory based on the concept of temporal coding of information. Here we show, however, that error-free retrieval states of such networks turn out to be typically unstable regardless of the network size, in contrast to the classical Hopfield model. We propose a remedy for this undesirable property, and provide a systematic study of the improved model. In particular, we show that the error-free capacity of the network is at least 2epsilon(2) / log n patterns per neuron, where n is the number of oscillators (neurons) and epsilon the strength of the second-order mode in the coupling function. (C) 2004 Elsevier B.V. All rights reserved.",
keywords = "neural networks, phase oscillators, random matrices, model, natural frequencies, synchronization, dynamics, cortex",
author = "Takashi Nishikawa and Hoppensteadt, {F C} and Ying-Cheng Lai",
year = "2004",
month = "10",
day = "1",
doi = "10.1016/j.physd.2004.06.011",
language = "English",
volume = "197",
pages = "134--148",
journal = "Physica. D, Nonlinear Phenomena",
issn = "0167-2789",
publisher = "Elsevier",
number = "1-2",

}

TY - JOUR

T1 - Oscillatory associative memory network with perfect retrieval

AU - Nishikawa, Takashi

AU - Hoppensteadt, F C

AU - Lai, Ying-Cheng

PY - 2004/10/1

Y1 - 2004/10/1

N2 - Inspired by the discovery of possible roles of synchronization of oscillations in the brain, networks of coupled phase oscillators have been proposed before as models of associative memory based on the concept of temporal coding of information. Here we show, however, that error-free retrieval states of such networks turn out to be typically unstable regardless of the network size, in contrast to the classical Hopfield model. We propose a remedy for this undesirable property, and provide a systematic study of the improved model. In particular, we show that the error-free capacity of the network is at least 2epsilon(2) / log n patterns per neuron, where n is the number of oscillators (neurons) and epsilon the strength of the second-order mode in the coupling function. (C) 2004 Elsevier B.V. All rights reserved.

AB - Inspired by the discovery of possible roles of synchronization of oscillations in the brain, networks of coupled phase oscillators have been proposed before as models of associative memory based on the concept of temporal coding of information. Here we show, however, that error-free retrieval states of such networks turn out to be typically unstable regardless of the network size, in contrast to the classical Hopfield model. We propose a remedy for this undesirable property, and provide a systematic study of the improved model. In particular, we show that the error-free capacity of the network is at least 2epsilon(2) / log n patterns per neuron, where n is the number of oscillators (neurons) and epsilon the strength of the second-order mode in the coupling function. (C) 2004 Elsevier B.V. All rights reserved.

KW - neural networks

KW - phase oscillators

KW - random matrices

KW - model

KW - natural frequencies

KW - synchronization

KW - dynamics

KW - cortex

U2 - 10.1016/j.physd.2004.06.011

DO - 10.1016/j.physd.2004.06.011

M3 - Article

VL - 197

SP - 134

EP - 148

JO - Physica. D, Nonlinear Phenomena

JF - Physica. D, Nonlinear Phenomena

SN - 0167-2789

IS - 1-2

ER -