ANALYSIS OF BOOLEAN LIKE TRAINING ALGORITHM (BLTA) FOR BINARY FEED FORWARD NEURAL NETWORK

Abstract

Artificial Neural R Networks (ANN) have been studied for the last three decades in the field of speech and image processing and for solving problems for which no algorithmic procedure exists. The study of various ways of building electronic networks which implemented the mathematical models of Biological neural networks, resulted in ANN. In recent years, the Back propagation training algorithm for the perceptron type neural network has been applied to many areas. In application areas that deals with the need for binary-to- binary mappings, the effectiveness of the back-propagation algorithm comes into question. ~.. 'J 1 •- ~f ii -' f U... Especially an extremely high number of iterations become necessary to possibly obtain even simple binary-to-binary mappings. A new training algorithm called the Boolean like training Algorithm (BLTA) is used to implement binary-to-binary mappings utilizing a four layer Binary feed forward Neural Network (BFNN), so firstly A Boolean like training algorithm is developed for 4 binary input/1 binary output function to be represented by BFNN. For this a neural network architecture is developed to carry out this work.