Szczegóły publikacji
Opis bibliograficzny
Selection of deep reinforcement learning using a genetic algorithm / Yurii Kryvenchuk, Dmytro Petrenko, Dariusz CICHOŃ, Yuriy Malynovskyy, Tetiana Helzhynska // W: COLINS 2022 [Dokument elektroniczny] : Computational Linguistics and Intelligent Systems : proceedings of the 6th international conference on Computational Linguistics and Intelligent Systems : Gliwice, Poland, May 12–13, 2022. Vol. 1, Main conference / ed. by Vasyl Lytvyn [et al.]. — Wersja do Windows. — Dane tekstowe. — [Aachen] : CEUR Workshop Proceedings, 2022. — (CEUR Workshop Proceedings ; ISSN 1613-0073 ; Vol. 3171). — S. [1129–1138]. — Wymagania systemowe: Adobe Reader. — Tryb dostępu: https://ceur-ws.org/Vol-3171/paper83.pdf [2022-12-01]. — Bibliogr. s. [1137–1138], Abstr.
Autorzy (5)
- Kryvenchuk Yurii
- Petrenko Dmytro
- AGHCichoń Dariusz
- Malynovskyy Yuriy
- Helzhynska Tetiana
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 143955 |
|---|---|
| Data dodania do BaDAP | 2022-12-21 |
| Rok publikacji | 2022 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Creative Commons | |
| Czasopismo/seria | CEUR Workshop Proceedings |
Abstract
Neural network models contain many parameters. The selection of these parameters and the selection of the correct model takes a long time. Model developers rely on expert assessment to select models and hyperparameters for them. This paper examines the use of a genetic algorithm as an alternative to the current design process. The genetic algorithm is used to automatically select network hyperparameters and the model of the network itself. This improves the network model and reduces development time. The general model presents an algorithm with input parameters equal to those required to represent possible states and system output parameters sufficient to describe possible actions. The algorithm automatically selects different models for different parameters. It is determined that the algorithm can successfully start working with a low-efficiency model template and show good model performance and adjust the indicators of the number of layers, policy, entropy coefficient, and others. This shows the potential for further application of these algorithms for drone design.
