采用时间差分算法的九路围棋机器博弈系统

A 9×9 Go computer game system using temporal difference

围棋机器博弈是机器博弈中重要的分支之一,其庞大的博弈空间给机器博弈研究者带来了巨大挑战.目前围棋机器博弈多采用静态估值搜索与蒙特卡洛树搜索,故将时间差分算法引入至九路围棋机器博弈系统中,提出基于时间差分算法的围棋机器博弈系统模型,该博弈系统具有一定的自学习能力,能在不断的对弈中逐步提高博弈能力.通过与采用α-β搜索算法的博弈系统进行实际对弈,证明了该方法的可行性.

Computer Go is an important branch of computer games and presents great challenges to computer game researchers due to its need for huge game space.Presently,the static evaluation method and the Monte-Carlo tree search method are widely used in Go computer games.In this paper,a temporal difference algorithm was introduced to the 9×9 Go computer game system which gave it self-learning capability,thereby improving the game levels as a result of the continuous training.Through playing chess with a system which adopts an α-β algorithm,the new method was proven to be effective.