Optimized joint NARX ANN - embedded processor design methodology

Author

Possignolo, Rafael Trapani ; Hammami, Omar

Author_Institution

ENSTA - ParisTech, Paris, France

fYear

2009

fDate

13-16 Dec. 2009

Firstpage

499

Lastpage

502

Abstract

Neural Networks are largely used in a vast number of applications, including time series prediction, function approximation, pattern classification. Recently Nonlinear Auto Regressive with eXogenous input (NARX) Recurrent Neural Networks has been used in to predict noisy and large time series (also referred as chaotic time series). This paper present a multiobjective optimized implementation of NARX neural network, specially designed to work on embedded systems.

Keywords

function approximation; logic design; optimisation; pattern classification; recurrent neural nets; time series; chaotic time series; embedded processor design; function approximation; noise prediction; nonlinear auto regressive with exogenous input; optimized joint NARX ANN; pattern classification; recurrent neural networks; time series prediction; Computer architecture; Design methodology; Design optimization; Embedded software; Entropy; Hardware; Neural networks; Neurons; Predictive models; Recurrent neural networks;

fLanguage

English

Publisher

ieee

Conference_Titel

Electronics, Circuits, and Systems, 2009. ICECS 2009. 16th IEEE International Conference on

Conference_Location

Yasmine Hammamet

Print_ISBN

978-1-4244-5090-9

Electronic_ISBN

978-1-4244-5091-6

Type

conf

DOI

10.1109/ICECS.2009.5410883

Filename

5410883