Skip to content

Do we need the num_labeled parameter for *_Supervised classes? #118

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

Closed
bellet opened this issue Aug 31, 2018 · 1 comment
Closed

Do we need the num_labeled parameter for *_Supervised classes? #118

bellet opened this issue Aug 31, 2018 · 1 comment
Milestone

Comments

@bellet
Copy link
Member

bellet commented Aug 31, 2018

Description

The *_Supervised classes allow to use a weakly supervised algorithm on labeled data by creating constraints from the labeled points.

I do not really understand why these classes have a parameter num_labeled, which is used to ignore some labeled points when creating constraints. To reduce the training complexity, it is enough to limit the number of constraints by using the num_constraints parameter. Using all available points to create the desired number of constraints can only benefit the algorithm as it sees more different points in the training constraints (i.e., it reduces variance).

I think num_labeled could be removed so that all available points are used to create the desired number of constraints.

@bellet
Copy link
Member Author

bellet commented Aug 31, 2018

Sorry I hit the create button by mistake, editing the issue description now

@bellet bellet changed the title Do we need the method random_subset of Constraints really useful ? Do we need the num_labeled parameter for *_Supervised classes? Aug 31, 2018
@perimosocordiae perimosocordiae added this to the v0.5.0 milestone Aug 31, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants