Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

About --relative-rank? #18

Open
mc-mong opened this issue Mar 2, 2024 · 2 comments
Open

About --relative-rank? #18

mc-mong opened this issue Mar 2, 2024 · 2 comments

Comments

@mc-mong
Copy link

mc-mong commented Mar 2, 2024

I'm curious about the --relative-rank option.
What about the -int- range?
Is it affected much by the -p(playouts) option?
Do you have an example you can refer to?

@CGLemon
Copy link
Owner

CGLemon commented Mar 3, 2024

The range of --relative-rank should be between 0~25. The higher value should be stronger and only work on the 19x19 game. Here is how it works.

We transform the relative-rank into the number of active children at the root node. The root will randomly select this number of children and prune other non-selected children. For example, the root randomly select 23 children at the first move if the relative-rank is 0.

The number of playouts option will significantly affect the strength, because the root may prune all high priority move. The policy network does not work in the case. The tree search could keep the quality of moves.

@mc-mong
Copy link
Author

mc-mong commented Mar 3, 2024

Thank you.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants