Dobrovsky, Aline; Borghoff, Uwe M.; Hofmann, Marko
Dokumenttyp:
Konferenzbeitrag / Conference Paper
Titel:
Deep Reinforcement Learning in Serious Games Analysis and Design of Deep Neural Network Architectures Eurocast 2017
Titel Konferenzpublikation:
16th International Conference on Computer Aided Systems Theory
Untertitel Konferenzpublikation:
Las Palmas De Gran Canaria, Spain, February 2017
Konferenztitel:
International Conference on Computer Aided Systems Theory (16., 2017, Las Palmas)
Konferenztitel:
Eurocast 2017
Tagungsort:
Las Palmas de Gran Canaria, Spain
Jahr der Konferenz:
2017
Datum Beginn der Konferenz:
19.02.2017
Datum Ende der Konferenz:
24.02.2017
Verlag:
IUCTC
Jahr:
2017
Seiten von - bis:
259-260
Sprache:
Englisch
Abstract:
Serious Games (SG) belong to the most important future e-learning trends; they attain enhanced public acceptance and importance. Although more frequently used in recruitment and training, their production is still effortful and expensive. The generation of human behaviour and general game playing remain prevalent trends and challenges. Serious games can profit from diverse behaviour to increase learning effectiveness and from general AI methods for easy adaption to different games. Deep reinforcement learning (DRL) offers an opportunity for application because it has shown considerable results and is widely applicable as a general method. DRL means the combination of reinforcement learning and deep learning methods. A famous example is deep Q-learning for learning to play Atari games, where a convolutional neural network was trained with a variant of Q-learning on different Atari games and partially outperformed human game players. «
Serious Games (SG) belong to the most important future e-learning trends; they attain enhanced public acceptance and importance. Although more frequently used in recruitment and training, their production is still effortful and expensive. The generation of human behaviour and general game playing remain prevalent trends and challenges. Serious games can profit from diverse behaviour to increase learning effectiveness and from general AI methods for easy adaption to different games. Deep reinforceme... »