General Type-2 Radial Basis Function Neural Network: A Data-Driven Fuzzy Model

Adrian Rubio-Solis, Patricia Melin, Uriel Martinez-Hernandez, George Panoutsos

Research output: Contribution to journalArticle

2 Citations (Scopus)
80 Downloads (Pure)

Abstract

This paper proposes a new General Type-2 Radial Basis Function Neural Network (GT2-RBFNN) that is functionally equivalent to a GT2 Fuzzy Logic System (FLS) of either Takagi-Sugeno-Kang (TSK) or Mamdani type. The neural structure of the GT2-RBFNN is based on the α-planes representation, in which the antecedent and consequent part of each fuzzy rule uses GT2 Fuzzy Sets (FSs). To reduce the iterative nature of the Karnik-Mendel algorithm, the Enhaned-Karnik-Mendel (EKM) type-reduction and three popular direct-defuzzification methods, namely the 1) Nie-Tan approach (NT), the 2) Wu-Mendel uncertain bounds method (WU) and the 3) Biglarbegian-Melek-Mendel algorithm (BMM) are used. Hence, this paper provides four different architectures of the GT2-RBFNN and their parametric optimisation. Such optimisation is a two-stage methodology that first implements an Iterative Information Granulation (IIG) approach to estimate the antecedent parameters of each fuzzy rule. Secondly, each consequent part and the fuzzy rule base of the GT2-RBFNN is optimised using an Adaptive Gradient Descent method (AGD) respectively. A number of popular benchmark data sets, the identification of a nonlinear system and the prediction of chaotic time series are considered. The reported comparative analysis of experimental results is used to evaluate the performance of the suggested GT2 RBFNN with respect to other popular methodologies.

Original languageEnglish
Article number8417444
Pages (from-to)333-347
Number of pages15
JournalIEEE Transactions on Fuzzy Systems
Volume27
Issue number2
Early online date23 Jul 2018
DOIs
Publication statusPublished - 1 Feb 2019

Fingerprint

Fuzzy rules
Neural networks
Granulation
Fuzzy sets
Fuzzy logic
Nonlinear systems
Time series
Identification (control systems)

Keywords

  • α-plane representation
  • Frequency selective surfaces
  • Fuzzy logic
  • fuzzy modelling
  • Fuzzy sets
  • General Type-2 FLSs
  • Iterative algorithms
  • Radial basis function networks
  • Radial Basis Function Neural Networks
  • Uncertainty

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Computational Theory and Mathematics
  • Artificial Intelligence
  • Applied Mathematics

Cite this

General Type-2 Radial Basis Function Neural Network: A Data-Driven Fuzzy Model. / Rubio-Solis, Adrian; Melin, Patricia; Martinez-Hernandez, Uriel; Panoutsos, George.

In: IEEE Transactions on Fuzzy Systems, Vol. 27, No. 2, 8417444, 01.02.2019, p. 333-347.

Research output: Contribution to journalArticle

Rubio-Solis, Adrian ; Melin, Patricia ; Martinez-Hernandez, Uriel ; Panoutsos, George. / General Type-2 Radial Basis Function Neural Network: A Data-Driven Fuzzy Model. In: IEEE Transactions on Fuzzy Systems. 2019 ; Vol. 27, No. 2. pp. 333-347.
@article{fd7b8980a1e54de9a868549c80b205a6,
title = "General Type-2 Radial Basis Function Neural Network: A Data-Driven Fuzzy Model",
abstract = "This paper proposes a new General Type-2 Radial Basis Function Neural Network (GT2-RBFNN) that is functionally equivalent to a GT2 Fuzzy Logic System (FLS) of either Takagi-Sugeno-Kang (TSK) or Mamdani type. The neural structure of the GT2-RBFNN is based on the α-planes representation, in which the antecedent and consequent part of each fuzzy rule uses GT2 Fuzzy Sets (FSs). To reduce the iterative nature of the Karnik-Mendel algorithm, the Enhaned-Karnik-Mendel (EKM) type-reduction and three popular direct-defuzzification methods, namely the 1) Nie-Tan approach (NT), the 2) Wu-Mendel uncertain bounds method (WU) and the 3) Biglarbegian-Melek-Mendel algorithm (BMM) are used. Hence, this paper provides four different architectures of the GT2-RBFNN and their parametric optimisation. Such optimisation is a two-stage methodology that first implements an Iterative Information Granulation (IIG) approach to estimate the antecedent parameters of each fuzzy rule. Secondly, each consequent part and the fuzzy rule base of the GT2-RBFNN is optimised using an Adaptive Gradient Descent method (AGD) respectively. A number of popular benchmark data sets, the identification of a nonlinear system and the prediction of chaotic time series are considered. The reported comparative analysis of experimental results is used to evaluate the performance of the suggested GT2 RBFNN with respect to other popular methodologies.",
keywords = "α-plane representation, Frequency selective surfaces, Fuzzy logic, fuzzy modelling, Fuzzy sets, General Type-2 FLSs, Iterative algorithms, Radial basis function networks, Radial Basis Function Neural Networks, Uncertainty",
author = "Adrian Rubio-Solis and Patricia Melin and Uriel Martinez-Hernandez and George Panoutsos",
year = "2019",
month = "2",
day = "1",
doi = "10.1109/TFUZZ.2018.2858740",
language = "English",
volume = "27",
pages = "333--347",
journal = "IEEE Transactions on Fuzzy Systems",
issn = "1063-6706",
publisher = "IEEE",
number = "2",

}

TY - JOUR

T1 - General Type-2 Radial Basis Function Neural Network: A Data-Driven Fuzzy Model

AU - Rubio-Solis, Adrian

AU - Melin, Patricia

AU - Martinez-Hernandez, Uriel

AU - Panoutsos, George

PY - 2019/2/1

Y1 - 2019/2/1

N2 - This paper proposes a new General Type-2 Radial Basis Function Neural Network (GT2-RBFNN) that is functionally equivalent to a GT2 Fuzzy Logic System (FLS) of either Takagi-Sugeno-Kang (TSK) or Mamdani type. The neural structure of the GT2-RBFNN is based on the α-planes representation, in which the antecedent and consequent part of each fuzzy rule uses GT2 Fuzzy Sets (FSs). To reduce the iterative nature of the Karnik-Mendel algorithm, the Enhaned-Karnik-Mendel (EKM) type-reduction and three popular direct-defuzzification methods, namely the 1) Nie-Tan approach (NT), the 2) Wu-Mendel uncertain bounds method (WU) and the 3) Biglarbegian-Melek-Mendel algorithm (BMM) are used. Hence, this paper provides four different architectures of the GT2-RBFNN and their parametric optimisation. Such optimisation is a two-stage methodology that first implements an Iterative Information Granulation (IIG) approach to estimate the antecedent parameters of each fuzzy rule. Secondly, each consequent part and the fuzzy rule base of the GT2-RBFNN is optimised using an Adaptive Gradient Descent method (AGD) respectively. A number of popular benchmark data sets, the identification of a nonlinear system and the prediction of chaotic time series are considered. The reported comparative analysis of experimental results is used to evaluate the performance of the suggested GT2 RBFNN with respect to other popular methodologies.

AB - This paper proposes a new General Type-2 Radial Basis Function Neural Network (GT2-RBFNN) that is functionally equivalent to a GT2 Fuzzy Logic System (FLS) of either Takagi-Sugeno-Kang (TSK) or Mamdani type. The neural structure of the GT2-RBFNN is based on the α-planes representation, in which the antecedent and consequent part of each fuzzy rule uses GT2 Fuzzy Sets (FSs). To reduce the iterative nature of the Karnik-Mendel algorithm, the Enhaned-Karnik-Mendel (EKM) type-reduction and three popular direct-defuzzification methods, namely the 1) Nie-Tan approach (NT), the 2) Wu-Mendel uncertain bounds method (WU) and the 3) Biglarbegian-Melek-Mendel algorithm (BMM) are used. Hence, this paper provides four different architectures of the GT2-RBFNN and their parametric optimisation. Such optimisation is a two-stage methodology that first implements an Iterative Information Granulation (IIG) approach to estimate the antecedent parameters of each fuzzy rule. Secondly, each consequent part and the fuzzy rule base of the GT2-RBFNN is optimised using an Adaptive Gradient Descent method (AGD) respectively. A number of popular benchmark data sets, the identification of a nonlinear system and the prediction of chaotic time series are considered. The reported comparative analysis of experimental results is used to evaluate the performance of the suggested GT2 RBFNN with respect to other popular methodologies.

KW - α-plane representation

KW - Frequency selective surfaces

KW - Fuzzy logic

KW - fuzzy modelling

KW - Fuzzy sets

KW - General Type-2 FLSs

KW - Iterative algorithms

KW - Radial basis function networks

KW - Radial Basis Function Neural Networks

KW - Uncertainty

UR - http://www.scopus.com/inward/record.url?scp=85050378524&partnerID=8YFLogxK

U2 - 10.1109/TFUZZ.2018.2858740

DO - 10.1109/TFUZZ.2018.2858740

M3 - Article

VL - 27

SP - 333

EP - 347

JO - IEEE Transactions on Fuzzy Systems

JF - IEEE Transactions on Fuzzy Systems

SN - 1063-6706

IS - 2

M1 - 8417444

ER -