Evolution of MiniMax Algorithm’s State Evaluation Heuristic for the Game of Abalone By Richard Wilson Dec 1, 2003 Main Topics • What is Abalone? • What are you evolving? • How did you do this? • Representation, Mutation, Reproduction • Fitness < or why this is slow > • The Next Step Abalone • Board game in the vein of chess & checkers • Ranked "Game of the decade" in 1998 • For ages 7 to 70 • WS2003 - CS347 – class project • An Indirect Source of much frustration for ws2003 An Abalone Board State Evaluation Heuristic • Estimate of moves till game over Implementation • ES to evolve the heuristic • Heuristic operators: – Number of pieces on the board – Relative arrangement – Pushable pieces – location ES • Board Matrix - Classical Genetic Algorithm ideas apply More different ES • Groupings, Endangered pieces, – Genetic Programming BEAGLE • http://www.gel.ulaval.ca/~beagle/ • An extensible evolutionary programming environment. Representation Mutation & Reproduction • Board location value matrix • Various other aspects – Groupings – Endangered pieces Fitness evaluation • Play the evolved candidate in a tournament. • Tournaments follow WS03-CS347 Ab_Net guidelines. • Type of tournament influences EA The Next Step • Evolution of predictive pruning techniques • Other games ( next semesters cs347 competition ) Questions?
Pages to are hidden for
"Wilson"Please download to view full document