Asynchronous computing and emergence of computational chaos

Jacob Barhen; Vladimir Protopopescu; Sarit Barhen; Jack Wells

doi:10.1016/S1474-6670(17)36413-3

Asynchronous computing and emergence of computational chaos

Jacob Barhen, Vladimir Protopopescu, Sarit Barhen, Jack Wells

Computational Sciences & Engineering Div

Research output: Contribution to journal › Conference article › peer-review

Abstract

One of the primary advantages of artificial neural networks is their inherent ability to perform massively parallel, nonlinear signal processing. However, if such networks must operate in a concurrently asynchronous regime, a phenomenon referred to as "computational chaos" may occur, which impedes the efficient retrieval of information usually stored in the system's aitractors. In this paper, we characterize the computational chaos occurring in a widely used neural network model by estimating the complete Lyapunov spectrum associated with us dynamics. We also provide conditions that prevent the emergence of computational chaos in such concurrently asynchronous neural networks.

Original language	English
Pages (from-to)	93-98
Number of pages	6
Journal	IFAC-PapersOnLine
Volume	36
Issue number	6
DOIs	https://doi.org/10.1016/S1474-6670(17)36413-3
State	Published - 2003
Event	2003 IFAC Conference on Analysis and Design of Hybrid Systems, ADHS 2003 - St Malo, Brittany, France Duration: Jun 16 2003 → Jun 18 2003

Funding

ACKNOWLEDGEME~TS. This research was performed in the Computer Science and Mathematics Division, Oak Ridge National Laboratory. Funding was provided by the Missile Defense Agency and by the DOE Otlice of Science under contract DE-AC05-OOOR22725 with UT - Battelle, LLC.

Funders	Funder number
DOE Otlice of Science	DE-AC05-OOOR22725
Missile Defense Agency

Keywords

Asynchronous computing
Computational chaos
Lyapunov spectrum
Neural networks

Access to Document

10.1016/S1474-6670(17)36413-3

Cite this

@article{3584e52befe2419f8531ad743f56882a,

title = "Asynchronous computing and emergence of computational chaos",

abstract = "One of the primary advantages of artificial neural networks is their inherent ability to perform massively parallel, nonlinear signal processing. However, if such networks must operate in a concurrently asynchronous regime, a phenomenon referred to as {"}computational chaos{"} may occur, which impedes the efficient retrieval of information usually stored in the system's aitractors. In this paper, we characterize the computational chaos occurring in a widely used neural network model by estimating the complete Lyapunov spectrum associated with us dynamics. We also provide conditions that prevent the emergence of computational chaos in such concurrently asynchronous neural networks.",

keywords = "Asynchronous computing, Computational chaos, Lyapunov spectrum, Neural networks",

author = "Jacob Barhen and Vladimir Protopopescu and Sarit Barhen and Jack Wells",

note = "Publisher Copyright: Copyright {\textcopyright} 2003 IFAC.; 2003 IFAC Conference on Analysis and Design of Hybrid Systems, ADHS 2003 ; Conference date: 16-06-2003 Through 18-06-2003",

year = "2003",

doi = "10.1016/S1474-6670(17)36413-3",

language = "English",

volume = "36",

pages = "93--98",

journal = "IFAC-PapersOnLine",

issn = "1474-6670",

publisher = "Elsevier",

number = "6",

}

TY - JOUR

T1 - Asynchronous computing and emergence of computational chaos

AU - Barhen, Jacob

AU - Protopopescu, Vladimir

AU - Barhen, Sarit

AU - Wells, Jack

PY - 2003

Y1 - 2003

N2 - One of the primary advantages of artificial neural networks is their inherent ability to perform massively parallel, nonlinear signal processing. However, if such networks must operate in a concurrently asynchronous regime, a phenomenon referred to as "computational chaos" may occur, which impedes the efficient retrieval of information usually stored in the system's aitractors. In this paper, we characterize the computational chaos occurring in a widely used neural network model by estimating the complete Lyapunov spectrum associated with us dynamics. We also provide conditions that prevent the emergence of computational chaos in such concurrently asynchronous neural networks.

AB - One of the primary advantages of artificial neural networks is their inherent ability to perform massively parallel, nonlinear signal processing. However, if such networks must operate in a concurrently asynchronous regime, a phenomenon referred to as "computational chaos" may occur, which impedes the efficient retrieval of information usually stored in the system's aitractors. In this paper, we characterize the computational chaos occurring in a widely used neural network model by estimating the complete Lyapunov spectrum associated with us dynamics. We also provide conditions that prevent the emergence of computational chaos in such concurrently asynchronous neural networks.

KW - Asynchronous computing

KW - Computational chaos

KW - Lyapunov spectrum

KW - Neural networks

UR - http://www.scopus.com/inward/record.url?scp=85064573772&partnerID=8YFLogxK

U2 - 10.1016/S1474-6670(17)36413-3

DO - 10.1016/S1474-6670(17)36413-3

M3 - Conference article

AN - SCOPUS:85064573772

SN - 1474-6670

VL - 36

SP - 93

EP - 98

JO - IFAC-PapersOnLine

JF - IFAC-PapersOnLine

IS - 6

T2 - 2003 IFAC Conference on Analysis and Design of Hybrid Systems, ADHS 2003

Y2 - 16 June 2003 through 18 June 2003

ER -

Asynchronous computing and emergence of computational chaos

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this