A fuzzy tolerating mechanism for the multivalued Neuron

Multi-valued Neuron with Periodic activation function (MVN-P) was proposed for solving classification problems. The boundaries between two distinct categories are precisely specified in MVN-P, which may cause slow convergence in learning or low classification accuracy in generalization. In this paper, we propose a revised model, MVN-PFT, in which a fuzzy tolerating buffer is provided around a boundary between two distinct categories. Incorrect assignments in the buffer can be tolerated in the training phase. Simulation results show that the revised model can learn faster and offer a higher classification accuracy than MVN-P.