Parallelized Backpropagation Neural Network Algorithm using Distributed System

Abstract

With rapid increase of computer system erformance during past decade, parallel processing becomes a crucial issue for omputer system. Different parallel computing based methods have been proposed in recent years for the development of system performance. In the present research work, back propagation neural network training algorithm has been parallelized using distributed environment. The neural network modelhas been trained and tested for the vibration analysis data. Parallel processing using back propagation neural network algorithm helps to increase the performance of the system and to decrease the convergence time for the training of the neural network. In this work a three-layer architecture of back propagation neural network is considered in distributed environment for the plate vibration problem. The input layer consists of two inputs, specified boundary condition and aspect ratio (m) of the elliptic plate. The output layer of the artificial neural network architecture consists of one output in the form of the corresponding frequency parameter (f). However, the number of nodes in the hidden layer has been taken as 10 and 15 for comparison the results. The parallelism of back propagation neural network has been trained and tested for the analysis of vibration data. The implementation of parallelism of back propagation neural network algorithm on a distributed computing system with good performance has been demonstrated. The convergence time for the training of back propagation neural network by parallel processing is faster as compared to the single processing. The application of the proposed research work gives a faster estimation for the analysis of vibration data. In this research work, parallel processing of back propagation neural network has achieved the performance of the system with desired accuracy of the results.