search

jonathan-laurent / AlphaZero.jl

A generic, simple and fast implementation of Deepmind's AlphaZero algorithm.
https://jonathan-laurent.github.io/AlphaZero.jl/stable/
MIT License
1.24k stars 140 forks source link

When exploring a position, what these abbreviations mean? #121

Closed StepHaze closed 2 years ago

StepHaze commented 2 years ago

When exploring a position, what these abbreviations mean:

Nmcts Vnet Pmcts Pnet UCT (I believe it's Upper Confidence bounds applied to Trees) Qmcts Qnet

StepHaze commented 2 years ago

Also what are Lv, Lp, Lreg, Linv, Hp, Hpnet ?

gwario commented 2 years ago

These abbreviations are the properties of the Monte Carlo Tree Search.... If you check out the paper, there is an explanation of the MCTS algorithm: https://www.nature.com/articles/nature24270

It will become clear.... N is the visit count, V the value (win/loss), ...