Anne Beyer

Yes, I also came across this [here](https://github.com/scikit-learn/scikit-learn/pull/31693#issuecomment-3632331775), and it should indeed be fixed separately. @lucyleeow let me know if you want to address it or if I should give it...

Starting to look into this, I have a general question: What is the reason for handling colors by different parameters in the binary (currently including multiclass `predict`) and multiclass cases?...

I tried it with the example code from the docs and noted the following: I guess the problem is that the data is truncated accordingly in `from_estimator`, but `xlim` and...

> What about in addition to setting xlim and ylim on the plot, we clip the data (for instance using `np.clip`)? @StefanieSenger I don't think this is useful here because...

I was thinking about whether there would be an obvious sanity check on the limits, but apart from the checks already implemented, I don't really see one right now. I...

@StefanieSenger I created an environment with python: 3.12.2 and matplotlib: 3.9.2, but when I check out the code from this PR, I still get the same [result as above](https://github.com/scikit-learn/scikit-learn/pull/31693#issuecomment-3490579615). Only...

> > that if you zoom into an area within one class, the background color changes to grey, > > Weird! We seem to specify colors so maybe this is...

I think I narrowed down the problem a bit more: When we limit the plotting area at the point suggested in this PR, we also calculate the responses only for...

On a related note (that I came across again while looking at https://github.com/scikit-learn/scikit-learn/pull/32867): There is a detailed handling of colors using the `multiclass_colors` parameter, but this is only taken into...

Thank you for pointing this out @peter-avg! This is somewhat related to #32565 (which originated from [this issue](https://github.com/scikit-learn/scikit-learn/issues/32394#issuecomment-3375647854)). While the underlying problem is the same (user passed class instead of...