Interval Type-2 Fuzzy Radial Basis Function Neuron Model for Neural Networks

Abstract

Computing with type-2 fuzzy sets (T2 FSs) may be considered highly computational due to the numerous arithmetic operations and type-reduction process, since it involves numerous embedded T2 FSs. To reduce the complexity, interval type-2 fuzzy sets (IT2 FSs) have been used. However, the type-reduction process of IT2 FSs may still require numerous computations. In this paper, we propose a simple interval type-2 fuzzy radial basis function neuron (IT2 RBFN model that can be incorporated in the structure of neural networks. As a result, fast computing neural network structures can be designed without a type-reduction process. Hence, our proposed neuron model can be used to reduce the computational load since the network would use only arithmetic operations of IT2 FSs and biases without the need of type-reduction.