Training of the recurrent neural networks for prediction

Savitsky, Jury; Golovko, Vladimir

Поиск по всему репозиторию:

Открыть/скачать файлы документа

Научный доклад (190.5Kb)

Автор

Savitsky, Jury
Golovko, Vladimir

Дата издания

1999

Издательство

BSUIR

Показать полное описание

Библиографическое описание

Savitsky, J. Training of the recurrent neural networks for prediction / J. Savitsky, V. Golovko // Pattern recognition and information processing : proceedings of Fifth International Cоnference, Minsk, 18–20 Мay 1999 / Belarusian state university of informatics and radioelectronics, Technical university of Szczecin [et. al] ; editors: Rauf Sadykhov [et al.]. –Minsk ; Szczecin, 1999. – Vol. 2. – P. 248–252 : il. – Bibliogr.: p. 252 (9 titles).

Аннотация

In this paper the technique of creation of effective methods of training recurrent neural network for prediction problems are discussed. The various functions of activation of neural units are considered. The adaptive algorithms of training of neural networks with varied functions of activation of neural elements are considered. The computing experiments on prediction of time series demonstrate possibilities of the developed methods.