A method of constructing fuzzy neural network based on rough set theory

• Published in: Proceedings of the 2003 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.03EX693) Vol. 3; pp. 1723 - 1728 Vol.3
• Main Authors: Xian-Ming Huang, Ji-Kai Yi, Yan-Hong Zhang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 2003
• Subjects: Arithmetic; Biological neural networks; Computer networks; Control engineering; Fuzzy control; Fuzzy logic; Fuzzy neural networks; Fuzzy set theory; Neural networks; Set theory
• DOI: 10.1109/ICMLC.2003.1259775
• ISBN: 0780378652; 9780780378650; 9780780381315; 0780381319

A method of constructing fuzzy neural network structure by using rough set theory is presented . Since rough set theory has strong ability of analyzing numerical value and fuzzy neural network has the ability of approximating function nicely, a neural network model which has good intelligibility, simple computation and fast convergence is constructed by combining both theory. The main process to construct this network is as follows: firstly to acquire rules from present data set by rough set theory; then the cell number of each layer and relevant initial parameters are constructed according to these rules; finally all kinds of parameters are computed by BP(back promulgation) arithmetic and the design of the network is finished. Also in this paper an example of approximating a 2D nonlinear function is discussed and the feasibility and validity of the method are proved.