KRILL HERD AND PIECEWISE-LINEAR INITIALIZATION ALGORITHMS FOR CONSTRUCTING TAKAGI-SUGENO SYSTEMS
I. A. Hodashinsky, I. V. Filimonenko, K. S. Sarin
Tomsk State University of Control Systems and Radioelectronics, prosp. Lenina 40, Tomsk, 634050
Keywords: инициализация, нечёткие системы, метаэвристические алгоритмы, аппроксимация, initialization, fuzzy systems, metaheuristic algorithms, approximation
Abstract
A method for designing Takagi-Sugeno fuzzy systems is proposed in which the structure of the systems is generated using a piecewise linear initialization algorithm, and parameters are optimized using a meta-heuristic algorithm. The obtained systems are tested against real data sets. The influence of some parameters of this algorithm on the approximation accuracy are analyzed. Estimates of the approximation accuracy and the number of fuzzy rules are compared with four known methods of design
|
