Influence of the Neural Network Hyperparameters on its Numerical Conditioning
In this paper, the task of assessment of numerical conditioning of multilayer perceptron, forecasting time series with sliding window method, has been considered. Performance of the forecasting perceptron with various hyperparameters sets, with different amount of neurons and various activation func...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | Russian |
| Published: |
Ministry of Education of the Republic of Belarus, Establishment The Main Information and Analytical Center
2020-04-01
|
| Series: | Цифровая трансформация |
| Subjects: | |
| Online Access: | https://dt.bsuir.by/jour/article/view/478 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832557142626795520 |
|---|---|
| author | S. V. Sholtanyuk |
| author_facet | S. V. Sholtanyuk |
| author_sort | S. V. Sholtanyuk |
| collection | DOAJ |
| description | In this paper, the task of assessment of numerical conditioning of multilayer perceptron, forecasting time series with sliding window method, has been considered. Performance of the forecasting perceptron with various hyperparameters sets, with different amount of neurons and various activation functions in particular, has been considered. Main factors, influencing on the neural net conditioning, have been revealed, as well as performance features, when using various activation functions. Formulas for assessment of condition numbers of individual components of the forecasting perceptron and of the neural network itself have been proposed. Comparative analysis of results of training the forecasting perceptron with various hyperparameters on modeled time series has been performed. Conditions, providing the best stability and conditioning for the neural network, have been formulated. |
| format | Article |
| id | doaj-art-0906811365194195a838746a8e5f9585 |
| institution | Kabale University |
| issn | 2522-9613 2524-2822 |
| language | Russian |
| publishDate | 2020-04-01 |
| publisher | Ministry of Education of the Republic of Belarus, Establishment The Main Information and Analytical Center |
| record_format | Article |
| series | Цифровая трансформация |
| spelling | doaj-art-0906811365194195a838746a8e5f95852025-02-03T05:39:02ZrusMinistry of Education of the Republic of Belarus, Establishment The Main Information and Analytical CenterЦифровая трансформация2522-96132524-28222020-04-01014350170Influence of the Neural Network Hyperparameters on its Numerical ConditioningS. V. Sholtanyuk0Belarusian State UniversityIn this paper, the task of assessment of numerical conditioning of multilayer perceptron, forecasting time series with sliding window method, has been considered. Performance of the forecasting perceptron with various hyperparameters sets, with different amount of neurons and various activation functions in particular, has been considered. Main factors, influencing on the neural net conditioning, have been revealed, as well as performance features, when using various activation functions. Formulas for assessment of condition numbers of individual components of the forecasting perceptron and of the neural network itself have been proposed. Comparative analysis of results of training the forecasting perceptron with various hyperparameters on modeled time series has been performed. Conditions, providing the best stability and conditioning for the neural network, have been formulated.https://dt.bsuir.by/jour/article/view/478time series forecastingneural networksperceptronnumerical conditioningactivation function |
| spellingShingle | S. V. Sholtanyuk Influence of the Neural Network Hyperparameters on its Numerical Conditioning Цифровая трансформация time series forecasting neural networks perceptron numerical conditioning activation function |
| title | Influence of the Neural Network Hyperparameters on its Numerical Conditioning |
| title_full | Influence of the Neural Network Hyperparameters on its Numerical Conditioning |
| title_fullStr | Influence of the Neural Network Hyperparameters on its Numerical Conditioning |
| title_full_unstemmed | Influence of the Neural Network Hyperparameters on its Numerical Conditioning |
| title_short | Influence of the Neural Network Hyperparameters on its Numerical Conditioning |
| title_sort | influence of the neural network hyperparameters on its numerical conditioning |
| topic | time series forecasting neural networks perceptron numerical conditioning activation function |
| url | https://dt.bsuir.by/jour/article/view/478 |
| work_keys_str_mv | AT svsholtanyuk influenceoftheneuralnetworkhyperparametersonitsnumericalconditioning |