Neural network segmentation of time seriesстатья

Статья опубликована в журнале из списка RSCI Web of Science

Информация о цитировании статьи получена из Web of Science
Статья опубликована в журнале из перечня ВАК
Статья опубликована в журнале из списка Web of Science и/или Scopus
Дата последнего поиска статьи во внешних источниках: 29 мая 2015 г.

Работа с статьей


[1] Neural network segmentation of time series / S. A. Dolenko, Y. V. Orlov, I. G. Persiantsev, Y. S. Shugai // Pattern Recognition and Image Analysis: Advances in Mathematical Theory and Applications. — 2003. — Vol. 13, no. 3. — P. 433–440. This paper focuses on the basic principles of using hierarchical neural network classifiers (HNC) for analysis of the dynamics of complex objects and determining structure. Recently, the authors proposed an algorithm for designing a HNC that combines supervised learning with self-organization. It was successfully applied to static pattern analysis. Then, it was further developed for analysis and segmentation of time series with switching dynamics. In this paper, we show that during HNC construction the original problem is decomposed into a set of problems of complexity adequate to the capacities of the neural network used in the nodes of the HNC. The size of a hidden layer of such a network is determined, which is sufficient for constructing a HNC capable of correctly classifying all of the patterns. Here, a further development of the algorithm is proposed which makes it possible to use it for analysis of time series with drift-type changing dynamics. The results of numerical experiments on the analysis of model pseudochaotic time series and of real biomedical and cosmophysical data are presented.

Публикация в формате сохранить в файл сохранить в файл сохранить в файл сохранить в файл сохранить в файл сохранить в файл скрыть