Performance Analysis of Transfer Functions in an Artificial Neural Network
R. Yogitha, G. Mathivanan · 2018
Artificial Neural Network (ANN) is one of the promising domains in Artificial Intelligence (AI). With the development of engineering and research in AI, there are so many real time applications that use ANN for various classification and regression problems. It is important to analyze the performance of a neural network to know which kind of neural network provides best result to a particular kind of application. The performance can be analyzed using various transfer functions, training algorithms and performance functions. This paper analyses the performance of various linear and non linear transfer functions used in the hidden and output layer of the neural network by applying these transfer functions to various levels of classification problems involving datasets of three levels of complexity. Performance of artificial neural network for classification can be assessed using training, validation and testing accuracy. Various training algorithms and performance functions are also applied to the ANN along with different transfer functions.