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.

Sign up for GitHub

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

Jump to bottom

Policy Surprise Weighting #36

Open
shindavid opened this issue Jan 30, 2023 · 0 comments
Open

Policy Surprise Weighting #36

shindavid opened this issue Jan 30, 2023 · 0 comments
Labels
KataGo replication task learning improvement

Comments

@shindavid
Copy link
Owner

Used by KataGo, described here.

Basically, weights the self-play-generated sample points based on how "surprising" they are. That is, if the MCTS-generated count-distribution looks very different from the policy-prior, then includes multiple copies of that row of data, so that the next generation neural network puts more weight on correcting it.

Implement this and validate value through experimentation.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
KataGo replication task learning improvement
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@shindavid