封闭域UCT算法的实现及其性能测试

UCT算法在9×9棋盘上的水平,已经接近最优秀的专业棋手,但在19×19棋盘上还不尽人意.一种可能的改进是用多个并行的局部UCT搜索来替代全局UCT搜索,以期用同样的计算资源来达到更深的有效搜索深度.本文力图沿着此研究思路做一些初步的尝试.我们通过改写Fuego的全局UCT搜索算法,本文实现了封闭域UCT搜索算法,并在64个封闭域围棋死活问题上进行了测试.初步结果表明,封闭域UCT算法能够在不使用任何领域知识的情况下,把给定问题的初始分支系数从5~16降低到2.3以下.从正确率和速度两项指标来看,多个并行的局部UCT搜索的研究思路具备可行性.

Implementation and Performance Evaluation of Fully-Enclosed-Region UCT Algorithm

While the playing performance of UCT algorithm is nearly the same as that of top professionals on a 9 ×9 Go board,its performance on 19×19 board still needs great improvements.One possible way is to use mul-tiple local UCT searches in parallel,which could use the same computing resources to reach deeper depth.This paper tries to do some tentative work on this direction.After modifying Fuego’s implementation of global UCT search algorithm,we implement a fully-enclosed-region UCT local search algorithm and test it by using 64 standard tsume go problems.Present results show that fully-enclosed region UCT can reduce original branching factors from 5—16 to 2. 3 without using any domain knowledge on Go.Judged by accuracy of testing results and computing speed,it looks that it is promising to do further research on running multiple local UCT searches in parallel.