Extended Kalman Filter of Training Feed Forward Artificial Neural Network

Abstract

Abstract In this paper discuss the use of artificial neural networks multilayer feed forward, which are training using and Extended Kalman Filter (EKF) algorithm one and training Back propagation algorithm standard way then compare them by using two types of data which are the original data represent time delay of the network (Aria phone) for telephone communications and simulation data. The results confirm that EKF approach is faster to the period time steps to train the neural networks compared to the second way.