8000 Allow other initialization strategies for IterativeImputer · Issue #22770 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

Allow other initialization strategies for IterativeImputer #22770

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
Open
eschibli opened this issue Mar 11, 2022 · 1 comment
Open

Allow other initialization strategies for IterativeImputer #22770

eschibli opened this issue Mar 11, 2022 · 1 comment
Labels
module:impute Needs Decision - Include Feature Requires decision regarding including feature New Feature

Comments

@eschibli
Copy link
Contributor

Describe the workflow you want to enable

I want to initialize an IterativeImputer using a KNNImputer rather than a SimpleImputer.

Describe your proposed solution

Modifying IterativeImputer to accept instances of _BaseImputer as initial_strategys should be fairly simple, though I'm aware it is under active development and I understand why this wouldn't be a priority. I'm happy to put together a pull request if it is likely to be accepted.

Describe alternatives you've considered, if relevant

I am not aware of any compelling alternatives.

Additional context

No response
@eschibli eschibli added Needs Triage Issue requires triage New Feature labels Mar 11, 2022
@glemaitre
Copy link
Member

I don't think that this is a good idea in practice.

I will refer to #21967 and the work of @aperezlebel

Combining IterativeImputer with KNNImputer will not scale since already both imputers do not scale well in terms of computation time. KNNImputer is also intractable with a large number of samples.

So before considering this feature, I would request a good benchmark to show that using a KNNImputer as initialization is actually useful in practice.

@glemaitre glemaitre added Needs Decision - Include Feature Requires decision regarding including feature and removed Needs Triage Issue requires triage labels Mar 15, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
module:impute Needs Decision - Include Feature Requires decision regarding including feature New Feature
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@cmarmo @glemaitre @eschibli
0