MATHEMATICAL MODELING
DATA ANALYSIS
INTELLIGENT SYSTEMS
D.Y. Nagornykh Teaching big hybrid neural networks for time series prediction
DISTRIBUTED SYSTEMS
REGULATORY FRAMEWORK OF AUTOMATED SYSTEMS SYNTHESIS
D.Y. Nagornykh Teaching big hybrid neural networks for time series prediction

Abstract.

The article describes hybrid approach in neural networks, used in prediction of time series, as well as the specific aspects of teaching hybrid neural networks consisting of Self-Organizing Maps (SOM) and Multilayer Perсeptron (MLP). Also, paper contains the results, gained during the process of building and teaching the large hybrid neural network and the new algorithm of equitable teaching for self-organizing layer.

Keywords:

neural nets, hybrid neural nets, self-organizing maps, time series prediction, function approximation.

PP. 54-61.

 REFERENCES

1. Ejov A., Shumskiy S. Neuro-computing and it's adoption in economics and business. – Moscow, 1998.
2. Osovskiy S. Neural nets for data processing. – Moscow: Finance and statistics, 2002. – p. 252-256
3. Afanasiev V.N. Time series analysis and prediction. – Moscow: Finance and statistics, 2010. – p. 224-241.
4. Dinh Nghia Do, Osowski S. Shape recognition using FFT preprocessing and neural network. Compel, 1998. –Vol. 17, No 5/6. -C. 658-666
5. Micel A., Efremova E. Data preprocessing methods for financial time series prediction systems. "Proceedings of Tomsk State University of Control Systems and Radioelectronics" № 3 (11), 2005.

2019 / 01
2018 / 04
2018 / 03
2018 / 02

© ФИЦ ИУ РАН 2008-2018. Создание сайта "РосИнтернет технологии".