Fuzzy Neural Networks as "Good" Function Approximators
Rita Lovassy, László Tamás Kóczy, Imre J. Rudas, László Gál · 2011
The paper discusses the generalization capability of two hidden layer neural networks based on various fuzzy operators introduced earlier by the authors as Fuzzy Flip-Flop based Neural Networks (FNNs), in comparison with standard networks tansig function based, MATLAB Neural Network Toolbox in the frame of a simple function approximation problem. Various fuzzy neurons, one of them based on new fuzzy intersection and union pair, and two other selected well known fuzzy operators (Łukasiewicz and Dombi operators) combined with standard negation had been proposed as suitable for the construction of novel FNNs. The experimental results show that these FNNs provide rather good generalization performance, with far better mathematical stability than the standardneural networks and are more suitable to avoid overfitting in the case of test data containing noisy items in the form of outliers.