How soft games can be played.
Philippe De Wilde · 1999
: This paper shows how games can be played that involve natural language, learning, and evolutionary fictitious play. This is implemented using respectively fuzzy logic, neural networks, and evolutionary programming. Because of this combination, they can be called soft games. The algorithm is described, and placed in the context of related research. KEYWORDS: intelligent agents, games, soft computing, learning, replicator dynamics, evolutionary dynamics 1 Intelligent agents and game theory An agent should not operate on its own. Communication is essential in a modern, distributed environment. Moreover, collaboration can have an effect greater than the sum of the individual contributions. No single ant, however sophisticated, can build a termite nest. In software terms this means that many relatively simple agents may be more powerful than a few sophisticated agents. This should not be taken to extremes. If agents become too simple, they will only be able to do straightforward operatio...