Identification of quasi-ARX neurofuzzy model with an SVR and GA approach

Yu Cheng, Lan Wang, Takayuki Furuzuki

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

The quasi-ARX neurofuzzy (Q-ARX-NF) model has shown great approximation ability and usefulness in nonlinear system identification and control. It owns an ARX-like linear structure, and the coefficients are expressed by an incorporated neurofuzzy (InNF) network. However, the Q-ARX-NF model suffers from curse-of-dimensionality problem, because the number of fuzzy rules in the InNF network increases exponentially with input space dimension. It may result in high computational complexity and over-fitting. In this paper, the curse-of-dimensionality is solved in two ways. Firstly, a support vector regression (SVR) based approach is used to reduce computational complexity by a dual form of quadratic programming (QP) optimization, where the solution is independent of input dimensions. Secondly, genetic algorithm (GA) based input selection is applied with a novel fitness evaluation function, and a parsimonious model structure is generated with only important inputs for the InNF network. Mathematical and real system simulations are carried out to demonstrate the effectiveness of the proposed method.

Original languageEnglish
Pages (from-to)876-883
Number of pages8
JournalIEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
VolumeE95-A
Issue number5
DOIs
Publication statusPublished - 2012 May

Fingerprint

Support Vector Regression
Neuro-fuzzy
Computational complexity
Identification (control systems)
Genetic algorithms
Genetic Algorithm
Function evaluation
Quadratic programming
Fuzzy rules
Model structures
Curse of Dimensionality
Nonlinear systems
Computational Complexity
Nonlinear System Identification
Model
Overfitting
Evaluation Function
System Simulation
Fitness Function
Fuzzy Rules

Keywords

  • Genetic algorithm
  • Input selection
  • Quasi-ARX neurofuzzy networks
  • Support vector regression

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Computer Graphics and Computer-Aided Design
  • Applied Mathematics
  • Signal Processing

Cite this

Identification of quasi-ARX neurofuzzy model with an SVR and GA approach. / Cheng, Yu; Wang, Lan; Furuzuki, Takayuki.

In: IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences, Vol. E95-A, No. 5, 05.2012, p. 876-883.

Research output: Contribution to journalArticle

@article{b9a439c53b0d40b8b90d8775a3154d90,
title = "Identification of quasi-ARX neurofuzzy model with an SVR and GA approach",
abstract = "The quasi-ARX neurofuzzy (Q-ARX-NF) model has shown great approximation ability and usefulness in nonlinear system identification and control. It owns an ARX-like linear structure, and the coefficients are expressed by an incorporated neurofuzzy (InNF) network. However, the Q-ARX-NF model suffers from curse-of-dimensionality problem, because the number of fuzzy rules in the InNF network increases exponentially with input space dimension. It may result in high computational complexity and over-fitting. In this paper, the curse-of-dimensionality is solved in two ways. Firstly, a support vector regression (SVR) based approach is used to reduce computational complexity by a dual form of quadratic programming (QP) optimization, where the solution is independent of input dimensions. Secondly, genetic algorithm (GA) based input selection is applied with a novel fitness evaluation function, and a parsimonious model structure is generated with only important inputs for the InNF network. Mathematical and real system simulations are carried out to demonstrate the effectiveness of the proposed method.",
keywords = "Genetic algorithm, Input selection, Quasi-ARX neurofuzzy networks, Support vector regression",
author = "Yu Cheng and Lan Wang and Takayuki Furuzuki",
year = "2012",
month = "5",
doi = "10.1587/transfun.E95.A.876",
language = "English",
volume = "E95-A",
pages = "876--883",
journal = "IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences",
issn = "0916-8508",
publisher = "Maruzen Co., Ltd/Maruzen Kabushikikaisha",
number = "5",

}

TY - JOUR

T1 - Identification of quasi-ARX neurofuzzy model with an SVR and GA approach

AU - Cheng, Yu

AU - Wang, Lan

AU - Furuzuki, Takayuki

PY - 2012/5

Y1 - 2012/5

N2 - The quasi-ARX neurofuzzy (Q-ARX-NF) model has shown great approximation ability and usefulness in nonlinear system identification and control. It owns an ARX-like linear structure, and the coefficients are expressed by an incorporated neurofuzzy (InNF) network. However, the Q-ARX-NF model suffers from curse-of-dimensionality problem, because the number of fuzzy rules in the InNF network increases exponentially with input space dimension. It may result in high computational complexity and over-fitting. In this paper, the curse-of-dimensionality is solved in two ways. Firstly, a support vector regression (SVR) based approach is used to reduce computational complexity by a dual form of quadratic programming (QP) optimization, where the solution is independent of input dimensions. Secondly, genetic algorithm (GA) based input selection is applied with a novel fitness evaluation function, and a parsimonious model structure is generated with only important inputs for the InNF network. Mathematical and real system simulations are carried out to demonstrate the effectiveness of the proposed method.

AB - The quasi-ARX neurofuzzy (Q-ARX-NF) model has shown great approximation ability and usefulness in nonlinear system identification and control. It owns an ARX-like linear structure, and the coefficients are expressed by an incorporated neurofuzzy (InNF) network. However, the Q-ARX-NF model suffers from curse-of-dimensionality problem, because the number of fuzzy rules in the InNF network increases exponentially with input space dimension. It may result in high computational complexity and over-fitting. In this paper, the curse-of-dimensionality is solved in two ways. Firstly, a support vector regression (SVR) based approach is used to reduce computational complexity by a dual form of quadratic programming (QP) optimization, where the solution is independent of input dimensions. Secondly, genetic algorithm (GA) based input selection is applied with a novel fitness evaluation function, and a parsimonious model structure is generated with only important inputs for the InNF network. Mathematical and real system simulations are carried out to demonstrate the effectiveness of the proposed method.

KW - Genetic algorithm

KW - Input selection

KW - Quasi-ARX neurofuzzy networks

KW - Support vector regression

UR - http://www.scopus.com/inward/record.url?scp=84860604073&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84860604073&partnerID=8YFLogxK

U2 - 10.1587/transfun.E95.A.876

DO - 10.1587/transfun.E95.A.876

M3 - Article

VL - E95-A

SP - 876

EP - 883

JO - IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences

JF - IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences

SN - 0916-8508

IS - 5

ER -