Monte Carlo Tree Search

撰写于 2018-09-07 修改于 2018-09-07 分类 Reinforcement Learning 标签 Reinforcement Learning

Monte Carlo Tree Search(MCTS) is a powerful method to generate optimal policies for AI. There are four steps in MCTS

  1. Seletion: select a best action in current state using UCB
  2. Expension: expand tree node
  3. Simulation: roll out to estimate the value of current state
  4. Update: update(backpropagate) parameters

Some helpful videos:


Site by GoingMyWay using Hexo & Random
备案号: 粤ICP备16087705号-1

I am a ML and RL research student