An Efficient Interval Type-2 Fuzzy Neural Network Based Classifier

In this paper, a simplified Karnik-Mendel (SKM) is proposed to realize type reduction for type-2 fuzzy models. To construct the fuzzy rule based neural network classifier, the clustering algorithm is employed to generate the nodes in the network, and we explain the interpretable construction process detailedly. In addition, to improve the classification performance, a long term memory based gradient optimization strategy is presented to optimize the structure parameters. Finally, the effectiveness of the proposed model EIT2FNN-C is verified by 2 benchmark classification data sets.