Comparing the efficiency of serial andparallel algorithms for training artificialneural networks using computer clusters

An estimation of the number of multiplication and addition operations fortraining articial neural networks by means of consecutive and parallel algorithmson a computer cluster is carried out. The evaluation of the eciency of thesealgorithms is developed. The multilayer perceptron, the Volterra network and thecascade-correlation network are used as structures of articial neural networks.Dierent methods of non-linear programming such as gradient and non-gradientmethods are used for the calculation of the weight coecients.[...]