-
-
Notifications
You must be signed in to change notification settings - Fork 25.1k
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
Make RidgeCV, LogisticRegressionCV, ... warn when found optimal regularization parameter lies at the boundary of the range #16398
Comments
Hi @ogrisel can I raise a pull request for this? |
Please do @divyaprabha123 ! I think it would be better to make 2 separate PRs: one to warn for such cases (which makes sense no matter the boundaries) and one to increase the default boundaries in a few models. |
Sure! I will do for both |
take |
@divyaprabha123 are you going to work on the other classes too, or only RidgeCV? In the first case, I've listed the ones that I think would also benefit from the change (in #16783). |
Yes I am! : ). @Reksbril |
I think I can finish off the work by @divyaprabha123 toward RidgeCV in #16408. Depending on how that goes I may have a look at the others listed by @Reksbril at #16783. |
I think the
RidgeCV().fit(X_train, y_train)
should warn the user if the found value foralpha_
is eitheralphas.min()
oralphas.max()
. E.g.We could add a new
boundary_warning=True
constructor parameter to make it possible to silence the warning.BTW, the default ranges could probably be extended between 1e-6 and 1e6 with 13 levels on the logspace whenever it is cheap to do so (e.g. for RidgeCV whos current default range much too narrow
[0.1, 1., 10.]
).The text was updated successfully, but these errors were encountered: