RESEACH OF OPPORTUNITIES OF PRACTICAL USE NEURAL NETWORKS FOR IDENTIFICATION OF THE BASIC PARAMETERS OF THE FRACTIONAL DIFFUSION (PREPRINT IBRAE-2002-12)

Preprint IBRAE-2002-12

Goloviznin V., Kiselev V., Korotkin I., Semenov V., Hromov A., Yurkov Y.

The inverse problem of identification of fractional diffusion parameters is considered according to practical measurements. For the decision of this problem it is offered to use neural networks. The brief review neural networks is given The various ways of representation of the information on an input in a neural network are considered. Results of research of noise influence and quantity of training examples on accuracy of definition of parameters fractional diffusion also are given. As results of practical measurements it is used numerical decisions of a direct problem.

Bibliographical reference 

Goloviznin V., Kiselev V., Korotkin I., Semenov V., Hromov A., Yurkov Y. RESEACH OF OPPORTUNITIES OF PRACTICAL USE NEURAL NETWORKS FOR IDENTIFICATION OF THE BASIC PARAMETERS OF THE FRACTIONAL DIFFUSION. (In Russian). Preprint № IBRAE-2002-12. Moscow: Nuclear Safety Institute RAS, 2002. 37 p. — Refs.: 6 items.