Skip to content

Conversation

@mathurinm
Copy link
Owner

No description provided.

@mathurinm mathurinm marked this pull request as draft November 3, 2021 08:53
warnings.filterwarnings("ignore", category=ConvergenceWarning)

supp = self.identify_support(X, y, lmbd, fit_intercept)
self.X = self.X[:, supp]
Copy link
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think if you do that here it will break because you will return a w/beta which is too small

Copy link
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

or you need to change the way get_result is implemented currently

@QB3
Copy link
Collaborator

QB3 commented Nov 18, 2021

glum is solving all problems at low precision but very slowly.
Is it due to the algorithm itself or max_iter is not passed correctly?
Capture d’écran de 2021-11-18 17-24-04

@QB3
Copy link
Collaborator

QB3 commented Nov 18, 2021

When I run glum on finance I have a memory error: this is logic dince it tries to instanciate a hessian.
I also a segmentation fault error when running glum on rcv1, do you think it is also a memory problem ?

@mathurinm
Copy link
Owner Author

@QB3 I think the "slow, exact" convergence should be posted on the benchopt/lasso and the segfault and hessian (separate issues IMO) on the glum repo, do you agree ?

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

Successfully merging this pull request may close these issues.

4 participants