Stable Kalman filter and neural network for the chaotic systems identification

Research output: Contribution to journalArticle

30 Citations (Scopus)

Abstract

© 2017 The Franklin Institute In this research, a modified Kalman filter is introduced for the adaptation of a neural network. The modified Kalman filter is an improved version of the extended Kalman filter based in the following two changes: (1) a term of the weights adaptation is modified in the modified algorithm to assure the uniform stability, convergence of the weights error, and local minimums avoidance, (2) the activation functions are used instead of the Jacobian terms in the modified algorithm to assure the boundedness of the weights error. The suggested algorithm is applied for the chaotic systems identification.
Original languageAmerican English
Pages (from-to)7444-7462
Number of pages19
JournalJournal of the Franklin Institute
DOIs
StatePublished - 1 Nov 2017

Fingerprint

Chaotic systems
System Identification
Kalman filters
Chaotic System
Kalman Filter
Identification (control systems)
Neural Networks
Neural networks
Uniform Stability
Activation Function
Extended Kalman filters
Term
Local Minima
Boundedness
Chemical activation

Cite this

@article{761559c9ccfe458b83b5c5d1fe113b11,
title = "Stable Kalman filter and neural network for the chaotic systems identification",
abstract = "{\circledC} 2017 The Franklin Institute In this research, a modified Kalman filter is introduced for the adaptation of a neural network. The modified Kalman filter is an improved version of the extended Kalman filter based in the following two changes: (1) a term of the weights adaptation is modified in the modified algorithm to assure the uniform stability, convergence of the weights error, and local minimums avoidance, (2) the activation functions are used instead of the Jacobian terms in the modified algorithm to assure the boundedness of the weights error. The suggested algorithm is applied for the chaotic systems identification.",
author = "{de Jes{\'u}s Rubio}, Jos{\'e}",
year = "2017",
month = "11",
day = "1",
doi = "10.1016/j.jfranklin.2017.08.038",
language = "American English",
pages = "7444--7462",
journal = "Journal of the Franklin Institute",
issn = "0016-0032",
publisher = "Elsevier Ltd",

}

TY - JOUR

T1 - Stable Kalman filter and neural network for the chaotic systems identification

AU - de Jesús Rubio, José

PY - 2017/11/1

Y1 - 2017/11/1

N2 - © 2017 The Franklin Institute In this research, a modified Kalman filter is introduced for the adaptation of a neural network. The modified Kalman filter is an improved version of the extended Kalman filter based in the following two changes: (1) a term of the weights adaptation is modified in the modified algorithm to assure the uniform stability, convergence of the weights error, and local minimums avoidance, (2) the activation functions are used instead of the Jacobian terms in the modified algorithm to assure the boundedness of the weights error. The suggested algorithm is applied for the chaotic systems identification.

AB - © 2017 The Franklin Institute In this research, a modified Kalman filter is introduced for the adaptation of a neural network. The modified Kalman filter is an improved version of the extended Kalman filter based in the following two changes: (1) a term of the weights adaptation is modified in the modified algorithm to assure the uniform stability, convergence of the weights error, and local minimums avoidance, (2) the activation functions are used instead of the Jacobian terms in the modified algorithm to assure the boundedness of the weights error. The suggested algorithm is applied for the chaotic systems identification.

U2 - 10.1016/j.jfranklin.2017.08.038

DO - 10.1016/j.jfranklin.2017.08.038

M3 - Article

SP - 7444

EP - 7462

JO - Journal of the Franklin Institute

JF - Journal of the Franklin Institute

SN - 0016-0032

ER -