An analytical approach to single node delay-coupled reservoir computing

Autor(en): Schumacher, J.
Toutounji, H.
Pipa, G. 
Stichwörter: Analytical approximation; Approximate analytical; Computational costs; Delay differential equations; Nonlinear computations; Orders of magnitude; Theoretical investigations; Time series prediction, Benchmarking; Differential equations, Neural networks
Erscheinungsdatum: 2013
Journal: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volumen: 8131 LNCS
Startseite: 26
Seitenende: 33
Zusammenfassung: 
Reservoir computing has been successfully applied in difficult time series prediction tasks by injecting an input signal into a spatially extended reservoir of nonlinear subunits to perform history-dependent nonlinear computation. Recently, the network was replaced by a single nonlinear node, delay-coupled to itself. Instead of a spatial topology, subunits are arrayed in time along one delay span of the system. As a result, the reservoir exists only implicitly in a single delay differential equation, numerical solving of which is costly. We derive here approximate analytical equations for the reservoir by solving the underlying system explicitly. The analytical approximation represents the system accurately and yields comparable performance in reservoir benchmark tasks, while reducing computational costs by several orders of magnitude. This has important implications with respect to electronic realizations of the reservoir and opens up new possibilities for optimization and theoretical investigation. © 2013 Springer-Verlag Berlin Heidelberg.
Beschreibung: 
Conference of 23rd International Conference on Artificial Neural Networks, ICANN 2013 ; Conference Date: 10 September 2013 Through 13 September 2013; Conference Code:99717
ISBN: 9783642407277
ISSN: 03029743
DOI: 10.1007/978-3-642-40728-4_4
Externe URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-84884922913&doi=10.1007%2f978-3-642-40728-4_4&partnerID=40&md5=10a26854cc4cdcb1bdf53eeeaa977be2

Show full item record

Page view(s)

3
Last Week
0
Last month
0
checked on Apr 23, 2024

Google ScholarTM

Check

Altmetric