The effect of delays on the performance of Layer Recurrent Network

Author

Li, Tey Ching ; Nordin, Farah Hani ; Yap, Keem Siah

Author_Institution

Dept. of Electron. & Commun. Eng., Univ. Tenaga Nasional, Kajang, Malaysia

fYear

2010

fDate

21-23 May 2010

Firstpage

1

Lastpage

4

Abstract

Layer Recurrent Network (LRN) is a dynamic network that has a feedback loop as well as a delay for each layer of the network except for the last layer. The main objective for this research is to study the effect of delays on the performance of the LRN in identifying a nonlinear model. A numerical experiment of the nonlinear model is set up before a set of input and output data is collected. The collected data is then used to train the LRN. The numbers of delays at the feedback loop is manipulated and the effect of the network performance is observed where it shows that the network has the best performance when the number of delay is set to more than the default/original value (which is one).

Keywords

feedforward neural nets; recurrent neural nets; delay; feedback loop; layer recurrent network; nonlinear model; Backpropagation algorithms; Delay effects; Feedback loop; MATLAB; Mathematical model; Mutual information; Neural networks; Neurons; Signal processing algorithms; Transfer functions; Layer Recurrent Network (LRN); delay; dynamic network;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing and Its Applications (CSPA), 2010 6th International Colloquium on

Conference_Location

Mallaca City

Print_ISBN

978-1-4244-7121-8

Type

conf

DOI

10.1109/CSPA.2010.5545274

Filename

5545274