Enhance "Choosing the Right Estimator" Graphic (scikit-learn algorithm cheat sheet)

Describe the issue linked to the documentation

In its user guide, scikit-learn offers a Choosing the right estimator which is an interactive scikit-learn algorithm cheat sheet that is great.

When thinking about new features for skore, I thought of enhancing the user guide and have a pedagogical table which, for each estimator, says:

• if it needs to be scaled,
• if it can handle categorical features,
• if it can handle missing data,
• if it holds some randomness (and where / why),
• if it can be paralleled,
• etc (full proper list to be determined).

EDIT:

• The scikit-learn graph / map is great, but not sufficient IMHO because I would like to have, for each estimator, if I need to normalize the data or not, etc -> guidelines for each estimator
• I would like a table that is separate from the map, this is also a cheat sheet but not to appear on the map, maybe at the bottom of the map on the same user guide page

When discussing this with @jeromedockes and @Vincent-Maladiere, they told me about scikit-learn's estimator tags such as is_regressor. It seems that that knowledge is already partially in the tags.

Suggest a potential alternative/fix

• Maybe scikit-learn could have a table in the user guide with guidelines for each estimator?
• Maybe scikit-learn could hold more tags? And the table could be built from those tags?

Start with the user guide's “Choosing the right estimator” page and the estimator tags documentation linked in the issue. Clarify the complete set of guidance fields and whether they should come from new tags before changing anything. Done requires an agreed table or tag design covering the selected estimator guidance and corresponding documentation.