Comparative Study of Universal Function Approximators (Neural Network, Fuzzy Logic, ANFIS) for Non-Linear Systems
Hamit Erdem, Ali Berkol, Mustafa Sert · 2015
The Fuzzy Logic (FL), Artificial Neural Networks (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) are known as universal function approximator which has been used in various applications. In general, a function approximator needs to select a function or a mapping algorithm among the well-defined methods that closely capture the input- output variables relation. This study compares the application of aforementioned artificial intelligence approximators by using two non-linear functions. The curve-fitting capability of approximators has been compared considering three main metrics. These metrics are; fitting accuracy (Root Mean Square Error (RMS)), memory occupation, program code size and running time. Additionally, the parameters which affect the performance of each system have been investigated in details. The entire analysis has been developed and accomplished by MATLAB.