An implementation of visualizing Monte Carlo Tree Search
Require: pip install pygame
.
Run run.py
to play.
在原作者基础上,整合了快手的wu-uct并行MCTS算法,达到同快手论文基本一致的效果。
It's a simplified version of an old Chinese chess.
- When chess A jump over chess B, B will be removed.
- The fewer chess left, the higher the score get.
Monte Carlo Tree Search (MCTS) is a classic planning algorithm applied in Alphago and AlphaZero. In this case, a pure MCTS (without value evaluation network) is used but still effective to solve the problem.
- Child nodes inherit the color of the parent node.
- Nodes at the same search depth are rendered in the same line.