Olivier Grisel

@lorentzenchr any preference for the naming of the "perfect" model option and legend? I have a slight preference for: - `plot_oracle=True` for the parameter, and "Oracle predictions" for the legend...

> "add a test to check the sample weight/repetition equivalence property" -> Not sure how to do that to be honest, even after checking linked comments. We can leave that...

> As for the color of the perfect line, by default it uses the second matplotlib colors from plt.rcParams["axes.prop_cycle"] (which is "#ff7f0e"), which I hard coded in the tests since...

> For the regression part, I'm not sure I've understood correctly whether I'm expected to do it, or whether it will be in a separate branch? That's fine with me,...

> Should I add this new display in `test_common_curve_display.py`? Indeed. That will help up check that we yield consistent error messages on invalid inputs in particular.

@JosephBARBIERDARNAL please also add a changelog entry under `whats_new/doc/upcoming_changes` so that we can get the CI global status of the PR green when the rest of the tests as green.

> For now, capcurve raises a NotImplementedError instead. I guess we don't want to test capcurve in this test because we want different behaviour for this display? I agree. Instead,...

> Does the example below correspond to what is required here? Yes, this is the kind of test I had in mind.

I merged `main` to be able to launch the new CUDA GPU CI workflow on this PR: - https://github.com/scikit-learn/scikit-learn/actions/runs/9387261905 EDIT: tests are green.

@betatim this is ready for a second review.