✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: f1e642f. Link to the linter CI: here}

@adrinjalali

I wouldn't change the digits dataset to cancer as it is not about accuracy, but the spirit of this example is to show a visualization of the importance (or relevance if you may) of each pixel when predicting a digit. I would rather merge the two examples (RFE and RFECV) as they are but adding a bit of narrative.

See my comment here on what the pixel color means and you can keep the introduction paragraph and the narrative on the dataset description from the original PR, as mentioned here.

@ArturoAmorQ I see, thanks for the clarification. I'll revert back to the digits dataset.

Given your suggestion to merge the RFE and RFECV examples, do we want to:

1. Show usage of RFE and RFECV, or
2. Show usage of just RFECV

@glemaitre has suggested in this comment that RFE and RFECV should be shown, so I'm going to go with that.

Since we are removing a file, we need to handle the redirection. Could you add the following in the redirects dictionary in doc/conf.py around l.300:

        "auto_examples/feature_selection/plot_rfe_breast_cancer"
    ),

I'll make sure to do this after I have finished updating the code 👍

I would rather merge the two examples (RFE and RFECV) as they are but adding a bit of narrative.

I am getting a bit confused and my comment for redirection is not anymore useful. @ArturoAmorQ do you think it could be a good idea to remove plot_rfe_with_cross_validation.py since it does not bring more value than the discussion that we will add here.

ping @ArturoAmorQ

do you think it could be a good idea to remove plot_rfe_with_cross_validation.py since it does not bring more value than the discussion that we will add here.

If we merge the two examples, yes, we can remove it. But the current version of RFECV displays error bars (which cannot be done with the heatmap using the digits dataset) and discusses in a simple dataset what happens with correlated features.

Notice that if we merge the two examples, that means using digits in a first section, and the synthetic dataset in a further section.

I think there are 2 slightly contradicting objectives here:

The first objective is to merge both examples because there's no point having 2 examples which both demonstrate the use of RFE/RFECV.

The second objective is to give the user:

• A visual example which shows how the RFE algorithm identifies important features, using the digits dataset
• A practical example which has a detailed narrative and analysis, using a synthetic dataset

A visual example which shows how the RFE algorithm identifies important features, using the digits dataset

A practical example which has a detailed narrative and analysis, using a synthetic dataset

We can use a single dataset for both aspect.

Another thing that we should do is to use an other estimator than SVM here since it does not scale with the number of samples. It would be best to use a LogisticRegression if we use a linear kernel.

@raj-pulapakura would you like to continue working on this PR?

Closing this PR as superseded by #28862.