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

consider scaling grad_norm by N #20

Open
ExpHP opened this issue Sep 21, 2017 · 2 comments
Open

consider scaling grad_norm by N #20

ExpHP opened this issue Sep 21, 2017 · 2 comments
Assignees
@ExpHP

Comments

@ExpHP
Copy link
Collaborator

ExpHP commented Sep 21, 2017 •
edited
Loading

note: the problem colin mentioned with doing this is that acgsd isn't supposed to know about the meaning or shape of the position vector (although it does already have grad_max which assumes that all elements are "alike" in meaning)
@colin-daniels
Copy link
Owner

Eh, I don't really have a problem with scaling it by N, but by 3N or something yeah.

@ExpHP
Copy link
Collaborator Author

ExpHP commented Sep 22, 2017

I got it: new option, we call it grad_rms_tol (rms = root mean square = norm / sqrt(num elements))

@ExpHP ExpHP self-assigned this Sep 22, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@ExpHP ExpHP

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@ExpHP @colin-daniels