A new training algorithm for HHFNN based on Gaussian membership function for approximation

A new training algorithm for hierarchical hybrid fuzzy–neural networks (HHFNN) based on Gaussian membership function is proposed in this paper. This new algorithm adjusts the widths of Gaussian membership functions of the IF parts of fuzzy rules in the lower fuzzy sub-systems, and updates the weights and bias terms of the upper neural network by gradient-descent method. Two advantages of the proposed algorithm are shown in this paper: firstly, its parameters are usually fewer, compared with the existing training algorithm for HHFNN and standard BP algorithm; secondly, it outperforms the latter two algorithms in accuracy according to simulation results.