Romanian Journal of Information Technology and Automatic Control / Vol. 17, No. 4, 2007

ALGORITM EVOLUTIV DE IDENTIFICARE NEURALĂ

Mihai Tertișco, George Condur

Abstract:

The paper presented an evolutionary algorithm guy concerning the neural model structure projection. Trough this algorithm calculated of the optimal nets structures in report with structure obtained trough the adhibition of the Kolmogorov theorem. Is proposed an empirical criterion of flapped the experimental data in 2 sets: set of neural net training and set for verification of tile neural model. Is present the examples and study of case. Is propose for training a hybrid algorithm of training result through the combinations of the backpropagation algorithm with the evolutional suggested algorithm of the author work.

Keywords:
Neural nets, neural models, feedforward, evolutionary algorithm, backpropagation.

View full article:

CITE THIS PAPER AS:
Mihai Tertișco, George Condur, "ALGORITM EVOLUTIV DE IDENTIFICARE NEURALĂ", Romanian Journal of Information Technology and Automatic Control, ISSN 1220-1758, vol. 17(4), pp. 23-34, 2007.