Tsukamoto-type neural fuzzy inference network

A Tsukamoto-type neural fuzzy inference network (TNFIN) is proposed. The TNFIN consists of a special five-layer feedforward neural fuzzy network. The fuzzy implication used in the paper is actually an inverse function transformation rather than the standard linguistic “if/then” rule. A hybrid learning algorithm combining the least square estimation method and the gradient descent method has been used to tune the parameters and speed up the learning process. To demonstrate the capability of the proposed TNFIN, two simulation examples (one in nonlinear function mapping and one in chaos time series prediction) are applied for validating the model. Simulation results show that the TNFIN model with less parameters and smaller iteration numbers produces the remarkable results