Fuzzy Neuron and Its Computational Capability

Based on analysis of the limitations of computational capability of a fuzzy neuron model, two improved methods on the definition of fuzzy neuron model are presented, by using non-monotonous function as restricted inspiring function and by modifying the definition domain. For the improved fuzzy neuron models, computational capability equivalence to Turing Machine is proved. The proposed models are proved to be generalizations of traditional neural networks. Finally, the characteristics of mix fuzzy neural networks are discussed.