Hardware Implementable Neural Networks Based on Fuzzy Operators

In this paper we propose sigmoid function generators derived from fuzzy J-K and D flip-flops, and Fuzzy Flip-Flop based Neural Networks (FNNs) based on Dombi, Łukasiewicz, and a new pair of Trigonometric operations. An advantage of such FNNs is their easy hardware implementability, and they are more suitable to avoid overfitting than standard neural networks (e.g. tansig function based, MATLAB Neural Network Toolbox type) in the frame of simple function approximation problems. The experimental results show that these FNNs provide rather good generalization performance, with far better mathematical stability than the standard tansig based neural networks, and they are more suitable to avoid overfitting in the case of test data containing noisy items in the form of outliers.