Automated Generation of New Concepts from General Game Playing

Abstract

AbstractIn this paper, we propose algorithms to extract explicit concepts from general games and these concepts are useful to understand semantics of games using General Game Playing as a research domain. General Game Playing is a research domain to invent game players which are able to play general games without any human intervention. There are many approaches to General Game Playing, for example, UCT, Neural Network, and Simulation-based approaches. Successful knowledge acquisition is reported in these approaches. However, generated knowledge is not explicit in conventional methods. We extract explicit concepts from heuristic functions obtained using a simulation based approach. Concepts to understand the semantics of Tic-tac-toe are generated by our approach. These concepts are also available to understand the semantics of Connect Four. We conclude that our approach is applicable to general games and is able to extract explicit concepts which are able to be understood by humans.KeywordsGeneral Game Playing (GGP)Heuristic FunctionSuccessful Knowledge AcquisitionPlayoutGame Description Language (GDL)These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.