Generalized type-2 fuzzy logic in response integration of modular neural networks

In this paper a new method for response integration, based on generalized type-2 fuzzy logic, in modular neural networks (MNNs) is presented. The main idea is that the uncertainty in combining the outputs of the different modules in the MNN can be handled in a better way by using type-2 fuzzy logic. Previous works have considered using interval type-2 fuzzy logic for this task, but in this paper we are proposing the use of generalized type-2 fuzzy logic to improve the overall results of MNNs. The new method was tested with the problem of face recognition, showing that generalized type-2 fuzzy logic outperforms other approaches for the same task.