Tutorial eBooks PDF File Download

Poker is an interesting test-bed for artificialintelligence research. It is a game of imperfectknowledge, where multiple competing agentsmust deal with risk management, opponentmodeling, unreliable information, and deception,much like decision-making applications in thereal world. Opponent modeling is one of themost difficult problems in decision-makingapplications and in poker it is essential toachieving high performance. This paperdescribes and evaluates the implicit and explicitlearning in the poker program L o k i. L o k iimplicitly “learns” sophisticated strategies byselectively sampling likely cards for theopponents and then simulating the remainder ofthe game. The program has explicit learning forobserving its opponents, constructing opponentmodels and dynamically adapting its play toexploit patterns in the opponents’ play. The resultis a program capable of playing reasonablystrong poker, but there remains considerableresearch to be done to play at a world-class level.