Approximation of 2D function using simplest neural networks — A comparative study and development of GUI system

Abstract

This paper demonstrates the function approximation capability of feed forward neural network (FFNN). An attempt has been made to develop the Graphical user Interface (GUI) system for function approximation. This GUI system can handle the function approximation of any nonlinear/linear function which can have any number of input variable and 6 output variables. Parameters of neural network can be set from a single panel. This GUI system provides approximation for various functions which made this GUI universal for the wide range of the users without theoretical knowledge about the function approximation. A Neural network with a single hidden layer has been used to approximate the functions. To train the neural network various type of learning algorithms has been accumulated in GUI system. Two 2-D benchmark problem has been tested. Finally comparison has been made which shows that Levenberg-Marquardt (LM) back propagation with single hidden layer FFNN converges faster than other training algorithms.