Thema4794: Unterschied zwischen den Versionen
Dh1659 (Diskussion | Beiträge) (Die Seite wurde neu angelegt: „{{Abschlussarbeit |Titel=Thinking Fast and Slow with Model-Based Reinforcement Learning |Abschlussarbeitstyp=Master |Betreuer=Mohammd Karam Daaboul |Partner=FZ…“) |
Dh1659 (Diskussion | Beiträge) |
||
Zeile 10: | Zeile 10: | ||
|Beschreibung DE=Reinforcement Learning (RL) is a method of machine learning in which an agent learns a strategy through interactions with its environment that maximizes the rewards it receives from the environment. Model-Based Reinforcement Learning (MBRL) algorithms can learn with notably fewer samples than other RL approaches by using a learned dynamics model of the environment. In this model, policy optimization is then performed (fast thinking), or the most optimal sequence of actions is deliberated (slow thinking). | |Beschreibung DE=Reinforcement Learning (RL) is a method of machine learning in which an agent learns a strategy through interactions with its environment that maximizes the rewards it receives from the environment. Model-Based Reinforcement Learning (MBRL) algorithms can learn with notably fewer samples than other RL approaches by using a learned dynamics model of the environment. In this model, policy optimization is then performed (fast thinking), or the most optimal sequence of actions is deliberated (slow thinking). | ||
− | The goal of this | + | The goal of this thesis is to design a hybrid approach to policy optimization and decision planning. For more information, see the pdf attached. |
Aktuelle Version vom 21. Juni 2021, 10:28 Uhr
Abschlussarbeitstyp: Master
Betreuer: Mohammd Karam Daaboul
Forschungsgruppe: Angewandte Technisch-Kognitive Systeme
Partner: FZI
Archivierungsnummer: 4794
Abschlussarbeitsstatus: Offen
Beginn:
21. Juni 2021
Abgabe: unbekannt
Reinforcement Learning (RL) is a method of machine learning in which an agent learns a strategy through interactions with its environment that maximizes the rewards it receives from the environment. Model-Based Reinforcement Learning (MBRL) algorithms can learn with notably fewer samples than other RL approaches by using a learned dynamics model of the environment. In this model, policy optimization is then performed (fast thinking), or the most optimal sequence of actions is deliberated (slow thinking).
The goal of this thesis is to design a hybrid approach to policy optimization and decision planning. For more information, see the pdf attached.
WIR BIETEN
- ein interdisziplinäres Forschungsumfeld mit Partnern aus Wissenschaft und Wirtschaft
- eine konstruktive Zusammenarbeit mit hellen, motivierten Mitarbeitern
- eine angenehme Arbeitsatmosphäre
WIR ERWARTEN
- Wissen auf dem Gebiet des Reinforcement Learning
- Fähigkeit sowohl State of the Art, als auch experimentelle Algorithmen zu implementieren
- Schnelle Auffassungsgabe (Schrittweise Einarbeitung in Tensorflow/PyTorch, Python/C++,)
- Sehr gute Deutsch- oder Englischkenntnisse
- Hohe Kreativität und Produktivität
ERFORDERLICHE UNTERLAGEN
- aktueller Notenauszug
- tabellarischer Lebenslauf
KONTAKT
M. Karam Daaboul
Ausschreibung: Download (pdf)