Comparison of three back-propagation training algorithms for two case studies

Abstract

This paper investigates the use of three back-propagation training algorithms, Levenberg-Marquardt, conjugate gradient and resilient back-propagation, for the two case studies, stream-flow forecasting and determination of lateral stress in cohesionless soils. Several neural network (NN) algorithms have been reported in the literature. They include various representations and architectures and therefore are suitable for different applications. In the present study, three NN algorithms are compared according to their convergence velocities in training and performances in testing. Based on the study and test results, although the Levenberg-Marquardt algorithm has been found being faster and having better performance than the other algorithms in training, the resilient back-propagation algorithm has the best accuracy in testing period.