D0CRT Accept Scikit-learn models directly instead of parameter dictionaries

[jpaillard]

• replace argument dicts by sklearn estimators
• remove the function _fit_lasso, replaced by cloned models
• Create a separate function lasso_screen for clarity
• add tests to see if the method works (identify truly important features) in a linear setting

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.08%. Comparing base (57d40de) to head (32b7b8d).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #370      +/-   ##
==========================================
- Coverage   98.10%   98.08%   -0.02%     
==========================================
  Files          22       22              
  Lines        1159     1148      -11     
==========================================
- Hits         1137     1126      -11     
  Misses         22       22              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[lionelkusch]

I think it's better to remove the screening procedure from D0CRT and add the screening function and the fit_function in another file.
What do you think?

[bthirion]

Thx. I think its' an improvement. Wondering if we can further simplify.

[jpaillard]

Would it actually make sense to have a Screening class instead of an independent function?
The class could not only be compatible with Lasso but also the univariate methods from #317

[lionelkusch]

Would it actually make sense to have a Screening class instead of an independent function? The class could not only be compatible with Lasso but also the univariate methods from #317

This can provide confusion between selection method and the Screening class.
We can add a marginal class using Lasso to determine the marginal features importance and use it to replace the actual screening process. In this case, any marginal method can be used before the importance of another method for making a preselection of features.

[bthirion]

Almost ready to merge, thx !

[bthirion]

If we provide no implementation of d1CRT, we should not make any claim on it.

[jpaillard]

I removed the last sentence

[lionelkusch]

Propose you small modification for improving the code.

[lionelkusch]

thank you for the change.

[bthirion]

LGTM, thx.