Skip to content
This repository has been archived by the owner on Jun 18, 2023. It is now read-only.

Expose "keep_regularized" argument for refit #74

Closed
valpasq opened this issue Dec 9, 2015 · 3 comments
Closed

Expose "keep_regularized" argument for refit #74

valpasq opened this issue Dec 9, 2015 · 3 comments
Milestone

Comments

@valpasq
Copy link
Collaborator

valpasq commented Dec 9, 2015

Need to expose keep_regularized argument for refit:

keep_regularized = TRUE
If lasso regression coefficient is fit to 0 in change detection stage, don't fit in refit stage.

keep_regularized = FALSE
Refit all coefficients in refit stage.

@ceholden ceholden added this to the v0.6.0 milestone Dec 9, 2015
@ceholden
Copy link
Owner

ceholden commented Dec 9, 2015

What do you think about renaming it to something along the lines of stay_regularized instead? I wrote the keyword argument keep_regularized and it even confuses me sometimes as to what that means (i.e., "keep" meaning retain the regularized coefficients or do we keep with the regularization that happened?).

@valpasq
Copy link
Collaborator Author

valpasq commented Dec 9, 2015

Yes! I like stay_regularized much better since we want the coefficients that were regularized to stay regularized, as opposed to keeping the coefficients themselves.

ceholden added a commit that referenced this issue Dec 15, 2015
@ceholden
Copy link
Owner

Implemented in commit referenced above.

refit_record function will retain keep_regularized kwarg for compatibility, but you can specify the behavior using stay_regularized in the config file. Just like the prefix, specify a list of True/False per refit method.

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

No branches or pull requests

2 participants