Details
| Originalsprache | Englisch |
|---|---|
| Titel des Sammelwerks | Proceedings of the 2018 IEEE Conference on Computational Intelligence and Games, CIG 2018 |
| Herausgeber (Verlag) | IEEE Computer Society |
| ISBN (elektronisch) | 9781538643594 |
| Publikationsstatus | Veröffentlicht - 11 Okt. 2018 |
| Extern publiziert | Ja |
| Veranstaltung | 14th IEEE Conference on Computational Intelligence and Games, CIG 2018 - Maastricht, Niederlande Dauer: 14 Aug. 2018 → 17 Aug. 2018 |
Publikationsreihe
| Name | IEEE Conference on Computatonal Intelligence and Games, CIG |
|---|---|
| Band | 2018-August |
| ISSN (Print) | 2325-4270 |
| ISSN (elektronisch) | 2325-4289 |
Abstract
This paper proposes a novel learning agent model for a General Video Game Playing agent. Our agent learns an approximation of the forward model from repeatedly playing a game and subsequently adapting its behavior to previously unseen levels. To achieve this, it first learns the game mechanics through machine learning techniques and then extracts rule-based symbolic knowledge on different levels of abstraction. When being confronted with new levels of a game, the agent is able to revise its knowledge by a novel belief revision approach. Using methods such as Monte Carlo Tree Search and Breadth First Search, it searches for the best possible action using simulated game episodes. Those simulations are only possible due to reasoning about future states using the extracted rule-based knowledge from random episodes during the learning phase. The developed agent outperforms previous agents by a large margin, while still being limited in its prediction capabilities. The proposed forward model approximation opens a new class of solutions in the context of General Video Game Playing, which do not try to learn a value function, but try to increase their accuracy in modelling the game.
ASJC Scopus Sachgebiete
- Informatik (insg.)
- Artificial intelligence
- Informatik (insg.)
- Computergrafik und computergestütztes Design
- Informatik (insg.)
- Maschinelles Sehen und Mustererkennung
- Informatik (insg.)
- Mensch-Maschine-Interaktion
- Informatik (insg.)
- Software
Zitieren
- Standard
- Harvard
- Apa
- Vancouver
- BibTex
- RIS
Proceedings of the 2018 IEEE Conference on Computational Intelligence and Games, CIG 2018. IEEE Computer Society, 2018. 8490411 (IEEE Conference on Computatonal Intelligence and Games, CIG; Band 2018-August).
Publikation: Beitrag in Buch/Bericht/Sammelwerk/Konferenzband › Aufsatz in Konferenzband › Forschung › Peer-Review
}
TY - GEN
T1 - Forward Model Approximation for General Video Game Learning
AU - Dockhorn, Alexander
AU - Apeldoorn, Daan
PY - 2018/10/11
Y1 - 2018/10/11
N2 - This paper proposes a novel learning agent model for a General Video Game Playing agent. Our agent learns an approximation of the forward model from repeatedly playing a game and subsequently adapting its behavior to previously unseen levels. To achieve this, it first learns the game mechanics through machine learning techniques and then extracts rule-based symbolic knowledge on different levels of abstraction. When being confronted with new levels of a game, the agent is able to revise its knowledge by a novel belief revision approach. Using methods such as Monte Carlo Tree Search and Breadth First Search, it searches for the best possible action using simulated game episodes. Those simulations are only possible due to reasoning about future states using the extracted rule-based knowledge from random episodes during the learning phase. The developed agent outperforms previous agents by a large margin, while still being limited in its prediction capabilities. The proposed forward model approximation opens a new class of solutions in the context of General Video Game Playing, which do not try to learn a value function, but try to increase their accuracy in modelling the game.
AB - This paper proposes a novel learning agent model for a General Video Game Playing agent. Our agent learns an approximation of the forward model from repeatedly playing a game and subsequently adapting its behavior to previously unseen levels. To achieve this, it first learns the game mechanics through machine learning techniques and then extracts rule-based symbolic knowledge on different levels of abstraction. When being confronted with new levels of a game, the agent is able to revise its knowledge by a novel belief revision approach. Using methods such as Monte Carlo Tree Search and Breadth First Search, it searches for the best possible action using simulated game episodes. Those simulations are only possible due to reasoning about future states using the extracted rule-based knowledge from random episodes during the learning phase. The developed agent outperforms previous agents by a large margin, while still being limited in its prediction capabilities. The proposed forward model approximation opens a new class of solutions in the context of General Video Game Playing, which do not try to learn a value function, but try to increase their accuracy in modelling the game.
KW - Belief Revision
KW - Breadth First Search
KW - Exception-tolerant Hierarchical Knowledge Bases
KW - Forward Model Approximation
KW - General Video Games
KW - Monte Carlo Tree Search
UR - http://www.scopus.com/inward/record.url?scp=85056837547&partnerID=8YFLogxK
U2 - 10.1109/CIG.2018.8490411
DO - 10.1109/CIG.2018.8490411
M3 - Conference contribution
AN - SCOPUS:85056837547
T3 - IEEE Conference on Computatonal Intelligence and Games, CIG
BT - Proceedings of the 2018 IEEE Conference on Computational Intelligence and Games, CIG 2018
PB - IEEE Computer Society
T2 - 14th IEEE Conference on Computational Intelligence and Games, CIG 2018
Y2 - 14 August 2018 through 17 August 2018
ER -