Dynamics of associative memory with a general transfer function

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

The Glauber dynamics of magnetic systems has been extended to the case of neural networks with a general odd response function. We have derived a set of recursion relations for the overlap parameter, noise average, and noise variance taken as macrovariables of the process describing the dynamics of associative memory. The retrieval process has been studied then for a hyperbolic tangent transfer function by the self-consistent signal to noise ratio method. The fatigue effect of the real neuron has been taken into account. The phase diagrams of the retrieval process reveal an enhanced storage capacity for a certain set of values of the parameters. Finally, a set of equations for the overlap parameters in the case of continuous asynchronous dynamics with nomonotone neurons has been analytically derived.

Original languageEnglish (US)
Pages (from-to)2619-2623
Number of pages5
JournalPhysical Review E
Volume51
Issue number3
DOIs
StatePublished - Jan 1 1995
Externally publishedYes

Fingerprint

associative memory
Associative Memory
transfer functions
Transfer Function
neurons
retrieval
Overlap
Neuron
Retrieval
Tangent function
Hyperbolic tangent
Odd function
Glauber Dynamics
Recursion Relations
Storage Capacity
Response Function
tangents
Fatigue
Phase Diagram
signal to noise ratios

ASJC Scopus subject areas

  • Statistical and Nonlinear Physics
  • Mathematical Physics
  • Condensed Matter Physics
  • Physics and Astronomy(all)

Cite this

Dynamics of associative memory with a general transfer function. / Opris, Ioan.

In: Physical Review E, Vol. 51, No. 3, 01.01.1995, p. 2619-2623.

Research output: Contribution to journalArticle

@article{7840d757e81f490b808140e9c923300a,
title = "Dynamics of associative memory with a general transfer function",
abstract = "The Glauber dynamics of magnetic systems has been extended to the case of neural networks with a general odd response function. We have derived a set of recursion relations for the overlap parameter, noise average, and noise variance taken as macrovariables of the process describing the dynamics of associative memory. The retrieval process has been studied then for a hyperbolic tangent transfer function by the self-consistent signal to noise ratio method. The fatigue effect of the real neuron has been taken into account. The phase diagrams of the retrieval process reveal an enhanced storage capacity for a certain set of values of the parameters. Finally, a set of equations for the overlap parameters in the case of continuous asynchronous dynamics with nomonotone neurons has been analytically derived.",
author = "Ioan Opris",
year = "1995",
month = "1",
day = "1",
doi = "10.1103/PhysRevE.51.2619",
language = "English (US)",
volume = "51",
pages = "2619--2623",
journal = "Physical review. E",
issn = "2470-0045",
publisher = "American Physical Society",
number = "3",

}

TY - JOUR

T1 - Dynamics of associative memory with a general transfer function

AU - Opris, Ioan

PY - 1995/1/1

Y1 - 1995/1/1

N2 - The Glauber dynamics of magnetic systems has been extended to the case of neural networks with a general odd response function. We have derived a set of recursion relations for the overlap parameter, noise average, and noise variance taken as macrovariables of the process describing the dynamics of associative memory. The retrieval process has been studied then for a hyperbolic tangent transfer function by the self-consistent signal to noise ratio method. The fatigue effect of the real neuron has been taken into account. The phase diagrams of the retrieval process reveal an enhanced storage capacity for a certain set of values of the parameters. Finally, a set of equations for the overlap parameters in the case of continuous asynchronous dynamics with nomonotone neurons has been analytically derived.

AB - The Glauber dynamics of magnetic systems has been extended to the case of neural networks with a general odd response function. We have derived a set of recursion relations for the overlap parameter, noise average, and noise variance taken as macrovariables of the process describing the dynamics of associative memory. The retrieval process has been studied then for a hyperbolic tangent transfer function by the self-consistent signal to noise ratio method. The fatigue effect of the real neuron has been taken into account. The phase diagrams of the retrieval process reveal an enhanced storage capacity for a certain set of values of the parameters. Finally, a set of equations for the overlap parameters in the case of continuous asynchronous dynamics with nomonotone neurons has been analytically derived.

UR - http://www.scopus.com/inward/record.url?scp=0000996049&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0000996049&partnerID=8YFLogxK

U2 - 10.1103/PhysRevE.51.2619

DO - 10.1103/PhysRevE.51.2619

M3 - Article

AN - SCOPUS:0000996049

VL - 51

SP - 2619

EP - 2623

JO - Physical review. E

JF - Physical review. E

SN - 2470-0045

IS - 3

ER -