Reduced-Set Vector-Based Interval Type-2 Fuzzy Neural Network

摘要

This paper describes an interval type-2 fuzzy modeling framework, reduced-set vector-based interval type-2 fuzzy neural network (RV-based IT2FNN), to characterize the representation in fuzzy logic inference procedure. The model proposed introduces the concept of interval kernel to interval type-2 fuzzy membership, and provides an architecture to extract reduced-set vectors for generating interval type-2 fuzzy rules. Thus, the overall RV-based IT2FNN can be represented as series expansion of interval kernel, and it does not have to determine the number of rules in advance. By using a hybrid learning mechanism, the proposed RV-based IT2FNN can construct an input-output mapping from the training data in the form of fuzzy rules. At last, simulation results show that the RV-based IT2FNN obtained possesses nice generalization and transparency.