RBFN to be used in adaptive fuzzy system (AFS). in common case; is assumed to be trained by means of the minimum necessary number of rules (hidden unit number) determination and adjusting of the mean and variance vectors of individual hidden nodes as well as their weights. In this paper, the simplest GP RBFN based adaptive fuzzy system for automatic fuzzy rule number determination is proposed (Fig.3). Only the network weights have been assumed to be adjusted by the GP algorithm while the centers and widths of unit sensitive zones were completely determined with the network input signal range and unit