Type-2 Fuzzy Classifier Ensembles for Text Entailment

This paper presents a new Type-2 Fuzzy Classifier ensemble, which enables to model parameter uncertainties by characterizing the fuzzy sets with secondary membership values. We use fuzzy clustering method to characterize primary membership values and genetic algorithm to approximate secondary membership grades. Furthermore, a weighing algorithm is used for a non-complex reduction for reasoning. We use transductive reasoning, instead of inductive reasoning, to develop a local model for every new vector, based on a nearness criterion vectors from the given database. It is shown that the method can improve classifier system modeling performance in comparison to well-known methods.