Alpha Go

Created
Created
2023 Dec 30 16:18
Editor
Creator
Creator
Seonglae ChoSeonglae Cho
Edited
Edited
2025 Mar 17 1:5
Monte Carlo Tree Search
because simulating space of whole game is too huge and there is a time limit for choosing action.
notion image
Use default policy for quick simulation. Default policy takes 1 micro second and tree policy takes 1 ms
Alpha go pre-trained based on human data and applied self-play with Efficient MCTS. Based on the result of self-play, it trains based on the result.
 
 
 
 
 

Recommendations