Reduced-set vector learning based on hybrid kernels for interval type 2 fuzzy modeling
Long Yu, Jian Bo Xiao, Song Wang · International Conference on Systems · 2008
This paper presents a new interval type-2 fuzzy inference system to handle uncertainty using reduced-set vector learning mechanism based on hybrid kernels. Firstly, a novel concept, interval kernel, is proposed. It establishes a relationship between interval type-2 fuzzy membership and hybrid kernel. According to it, a particular interval type-2 fuzzy inference system is built, which abandons traditional type reduction procedure and utilizes directly defuzzification after inference. Subsequently, the model optimization is realized via a hybrid learning mechanism involving two sub-algorithms: bottom-up simplification algorithm and quadratic programming combined with back propagation algorithm. At last, simulation results show that the interval type-2 fuzzy model obtained possesses nice generalization and transparency.