Розвиток паралельних методів та алгоритмів розв’язання задач цифрової фільтрації
The quasisystolic method of organizing computations has been progressed for solving filtering problems using adaptive smoothing and multiple cascade digital filtering. This method differs from the purely systo-lic one in that it allows data transmission from one port to several reception points at o...
Gespeichert in:
| Datum: | 2026 |
|---|---|
| Hauptverfasser: | , , |
| Format: | Artikel |
| Sprache: | Ukrainisch |
| Veröffentlicht: |
Інститут прикладних проблем механіки і математики ім. Я. С. Підстригача НАН України
2026
|
| Schlagworte: | |
| Online Zugang: | https://www.fmmit.lviv.ua/index.php/fmmit/article/view/433 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | Physico-mathematical modeling and informational technologies |
| Завантажити файл: | |
Institution
Physico-mathematical modeling and informational technologies| Zusammenfassung: | The quasisystolic method of organizing computations has been progressed for solving filtering problems using adaptive smoothing and multiple cascade digital filtering. This method differs from the purely systo-lic one in that it allows data transmission from one port to several reception points at once. Based on the mentioned method, optimal by speed parallel-pipeline algorithms for digital filtering of distorted data ha-ve been constructed. Optimality is proven in the specified classes of algorithms, which are equivalent in terms of the information graph. These algorithms are oriented towards implementation on quasisystolic computational structures and computers with a structural-procedural organization of calculations. The si-milarity of the structure of computations according to these filtering algorithms and computations perfor-med in convolutional neural networks gives grounds for using the latter when developing a more universal filter for preprocessing data of various types. For this purpose, a hybrid neural network architecture is proposed – a convolutional autoencoder. Numerical experiments have confirmed the high efficiency of such a filter. The obtained scientific results can be used in the pre-processing of large data arrays in many subject areas using modern computing means. |
|---|---|
| DOI: | 10.15407/fmmit2026.42.058 |