150 likes | 265 Views
Genetically Programmed Strategies For Chess Endgame. Introduction. Backgroud Evolving Strategies For KRK Results. Background. Computing Brute-Force Algorithm Pattern Use DataBase instead of BFA Unstandable Readable Alpha-Beta pruning Algorithm Improve evalution function
E N D
Genetically Programmed Strategies For Chess Endgame
Introduction • Backgroud • Evolving Strategies For KRK • Results
Background • Computing • Brute-Force Algorithm • PatternUse DataBase instead of BFA • Unstandable Readable • Alpha-Beta pruning Algorithm • Improve evalution function • Genetic Algorithm by neural network
Evolving Strategies For KRK • Focus on tree of moves and evalution functions not algorithm and not table • Learn how to play effectively use patten • Use Genetic Programming (GP) • Compute Correct Strategy • Generate an effective strategy with fitness function (outcome)
Evolving Strategies For KRK • Pattern:good enough for strategy • Genetic encoding and genetic operator • Binary Tree with operator-strategy • Operator as Node; Function(moves) as Leaves • Each Operator can be read use if-then-else • Transition(node-node) use logical AND • Mutation(replace operator by random)
Evolving Strategies For KRK • Fitness: • fi:partical better game value • Mi:total moves • N:program tree size
Results • First experiment
Results • Fourth experiment
Results • Compute Strategy • Pattern