On the Realization of Back-Propagation on a Transputer Based System

Abstract

This paper presents some experimental results on the realization of a parallel simulation of an Artificial Neural Network (ANN) on a Transputer based system. The Back-Propagation algorithm has been used, and the parallelization technique utilized is the Processor Farm based on the partition of the training patterns between the available transputers so that each transputer (trainee) is trained on a different part of the training set, while another transputer (the controller) keeps track of the global network state