Evolutionary Learning for Neuro-fuzzy Ensembles with Generalized Parametric Triangular Norms

摘要

In this paper we present a method for designing neuro-fuzzy systems with Mamdani-type inference and parametric t-norm connecting rule antecedents. Hamacher product was used as t-norm. The neuro-fuzzy systems are used to create an ensemble of classifiers. After obtaining the ensemble by bagging, every neuro-fuzzy system has its t-norm parameters fine-tuned. Thanks to this the accuracy is improved and the number of parameters can be reduced. The proposed method is tested on a well known benchmark.