PROJECT TITLE :

Multiple Opponent Optimization of Prisoner’s Dilemma Playing Agents

ABSTRACT:

Agents for playing iterated prisoner's dilemma are commonly trained using a coevolutionary system in that a player's score against a selection of other members of an evolving population forms the fitness perform. During this study we tend to examine instead a version of evolutionary iterated prisoner's dilemma in which an agent's fitness is measured as the common score it obtains against a fixed panel of opponents known as an examination board. The performance of agents trained using examination boards is compared against agents trained in the same old coevolutionary fashion. This includes assessing the relative competitive ability of players evolved with evolution and coevolution. The problem of many experimental boards as optimization problems is compared. A number of new sorts of methods are introduced. These include sugar ways which can be exploited with some difficulty and treasure hunt strategies that have multiple trapping states with completely different levels of exploitability. The degree to which ways trained with completely different examination boards produce different agents is investigated using fingerprints.