Alpha-beta search on virtual tree machines

Abstract Alpha-beta search and other tree-searching algorithms are important in the field of artificial intelligence, in particular for game-playing programs. Parallel processing methods may be used to search larger trees than would otherwise be possible, although there is an overhead involved when parallelism is introduced. We present an algorithm for deciding where and when to solve subproblems on a fixed finite network of processors, suitable for search of ordered game trees. Using this algorithm, search of large and irregular trees may be made on a much smaller physical network. The system proposed has been simulated, and we give results obtained in practice for one particular problem.