| |
|
Reinforcement Learning for Bomberman Type Game
Adamčiak, Jakub ; Beran, Vítězslav (referee) ; Hradiš, Michal (advisor)
This bachelor's thesis aims to develop, implement and train reinforcement learning models for a Bomberman-type game. It is based on Bomberland environment from CoderOne. This environment was created for education and research in the field of artificial intelligence. In this thesis I tackle the settings and problems of implementing agent into the environment. I used 2 policies (MLP and CNN), 2 algorithms (PPO and A2C) and 5 setups of neural networks for feature extraction with the use of libraries stable baselines 3 and pytorch. Total training time resulted in 1207 real-world hours, 4168 computing hours and 271 milions of time steps. Although the training was not successful, this thesis shows the process of implementing a reinforcement learning model into a Gym environment.
|
|
Combined heat and power production planning in a waste-to-energy plant using machine learning
Kollmann, Marek ; Miklas, Václav (referee) ; Touš, Michal (advisor)
V rámci tohoto výzkumu bylo použito strojové učení k optimalizaci plánování výroby na den dopředu zařízení na energetické zpracování odpadu (Waste-to-Energy, WtE), které se potýká s problémy, jako jsou nekvalitní data, nekontrolovatelná externí spotřeba a kolísající výroba páry v důsledku použití odpadu jako zdroje paliva. Hlavním cílem bylo předpovídat s vysokou přesností výkon přenášený do sítě, čehož bylo dosaženo vytvořením komplexního modelu sestávajícího ze sedmi dílčích kaskádovitě uspořádaných modelů. Každý dílčí model byl kriticky vyhodnocen pomocí standardních ukazatelů, jako je R2 a průměrná absolutní chyba. Zjištění odhalila významné zlepšení přesnosti předpovědí, což vedlo k vyváženějším výrobním plánům a snížení provozních penále. Tento přístup vedl k odhadovanému ročnímu zvýšení dodaného výkonu o 13 % a zisku o 2,6 milionu Kč pro konkrétní závod.
|
|
Anomaly Detection by IDS Systems
Gawron, Johann Adam ; Homoliak, Ivan (referee) ; Očenášek, Pavel (advisor)
The goal of this thesis is to familiarize myself, and the reader, with the issues surrounding anomaly detection in network traffic using artificial inteligence. To propose and subsequently implement a methodology for creating an anomaly classifier for network communication profiles. The classification method should be able to efficiently and accurately identify anomalies in network traffic to avoid generating false outputs. During the research of the issue, IDS systems, various types of attacks, and approaches to anomaly detection and classification were examined. In evaluating the effectiveness, several standard methods were examined and used to express the quality of classifiers.
|
|
SVM Algorithm Training for DDoS on SDN Networks
Murtadha ; Shujairiand ; Škorpil, Vladislav
Despite the flexibility provided by SDN technology is also vulnerable to attacks such as DDoS attacks, Network DDoS attack is a serious threat to the Internet today because internet traffic is increasing day by day, it is difficult to distinguish between legitimate and malicious traffic. To alleviate the DDoS attack in the campus network, to mitigate this attack, propose in this paper to classify benign traffic from DDoS attack traffic by SVM of the classification algorithms based on machine learning. As the contribution of this paper is to train the SVM algorithmwhich has been used in the approach for the training process. Due to the complexity of the dataset, using a type of kernel called a polynomial kernel to accomplish non-linearity discriminative. The results showed that the traffic classification was with the highest accuracy 96 %.
|
|
Reinforcement Learning for Bomberman Type Game
Adamčiak, Jakub ; Beran, Vítězslav (referee) ; Hradiš, Michal (advisor)
This bachelor's thesis aims to develop, implement and train reinforcement learning models for a Bomberman-type game. It is based on Bomberland environment from CoderOne. This environment was created for education and research in the field of artificial intelligence. In this thesis I tackle the settings and problems of implementing agent into the environment. I used 2 policies (MLP and CNN), 2 algorithms (PPO and A2C) and 5 setups of neural networks for feature extraction with the use of libraries stable baselines 3 and pytorch. Total training time resulted in 1207 real-world hours, 4168 computing hours and 271 milions of time steps. Although the training was not successful, this thesis shows the process of implementing a reinforcement learning model into a Gym environment.
|
| |