Spike-based cross-entropy method for reconstruction

A. Lőrincz, Zsolt Palotai, Gábor Szirtes

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

Most neural optimization algorithms use either gradient tuning methods or complicated recurrent dynamics that may lead to suboptimal solutions or require huge number of iterations. Here we propose a framework based on the cross-entropy method (CEM). CEM is an efficient global optimization technique, but it requires batch access to many samples. We transcribed CEM to an online form and embedded it into a reconstruction network that finds optimal representations in a robust way as demonstrated by computer simulations. We argue that this framework allows for neural implementation and suggests a novel computational role for spikes in real neuronal systems.

Original languageEnglish
Pages (from-to)3635-3639
Number of pages5
JournalNeurocomputing
Volume71
Issue number16-18
DOIs
Publication statusPublished - Oct 2008

Fingerprint

Entropy
Global optimization
Tuning
Computer Simulation
Computer simulation

Keywords

  • Cross-entropy method
  • Reconstruction networks
  • Spike-based reconstruction

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Cognitive Neuroscience

Cite this

Spike-based cross-entropy method for reconstruction. / Lőrincz, A.; Palotai, Zsolt; Szirtes, Gábor.

In: Neurocomputing, Vol. 71, No. 16-18, 10.2008, p. 3635-3639.

Research output: Contribution to journalArticle

Lőrincz, A. ; Palotai, Zsolt ; Szirtes, Gábor. / Spike-based cross-entropy method for reconstruction. In: Neurocomputing. 2008 ; Vol. 71, No. 16-18. pp. 3635-3639.
@article{6e87de6382e94bb0809328d27313b76d,
title = "Spike-based cross-entropy method for reconstruction",
abstract = "Most neural optimization algorithms use either gradient tuning methods or complicated recurrent dynamics that may lead to suboptimal solutions or require huge number of iterations. Here we propose a framework based on the cross-entropy method (CEM). CEM is an efficient global optimization technique, but it requires batch access to many samples. We transcribed CEM to an online form and embedded it into a reconstruction network that finds optimal representations in a robust way as demonstrated by computer simulations. We argue that this framework allows for neural implementation and suggests a novel computational role for spikes in real neuronal systems.",
keywords = "Cross-entropy method, Reconstruction networks, Spike-based reconstruction",
author = "A. Lőrincz and Zsolt Palotai and G{\'a}bor Szirtes",
year = "2008",
month = "10",
doi = "10.1016/j.neucom.2008.03.007",
language = "English",
volume = "71",
pages = "3635--3639",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "16-18",

}

TY - JOUR

T1 - Spike-based cross-entropy method for reconstruction

AU - Lőrincz, A.

AU - Palotai, Zsolt

AU - Szirtes, Gábor

PY - 2008/10

Y1 - 2008/10

N2 - Most neural optimization algorithms use either gradient tuning methods or complicated recurrent dynamics that may lead to suboptimal solutions or require huge number of iterations. Here we propose a framework based on the cross-entropy method (CEM). CEM is an efficient global optimization technique, but it requires batch access to many samples. We transcribed CEM to an online form and embedded it into a reconstruction network that finds optimal representations in a robust way as demonstrated by computer simulations. We argue that this framework allows for neural implementation and suggests a novel computational role for spikes in real neuronal systems.

AB - Most neural optimization algorithms use either gradient tuning methods or complicated recurrent dynamics that may lead to suboptimal solutions or require huge number of iterations. Here we propose a framework based on the cross-entropy method (CEM). CEM is an efficient global optimization technique, but it requires batch access to many samples. We transcribed CEM to an online form and embedded it into a reconstruction network that finds optimal representations in a robust way as demonstrated by computer simulations. We argue that this framework allows for neural implementation and suggests a novel computational role for spikes in real neuronal systems.

KW - Cross-entropy method

KW - Reconstruction networks

KW - Spike-based reconstruction

UR - http://www.scopus.com/inward/record.url?scp=56549104376&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=56549104376&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2008.03.007

DO - 10.1016/j.neucom.2008.03.007

M3 - Article

VL - 71

SP - 3635

EP - 3639

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 16-18

ER -