Christian Lorentzen

But `raw_predict` without something like `response_function`/`inverse` is quite meaningless. Maybe, we don't need the `link` function itself, only it's inverse. EDIT: Ok, starting with only `raw_predict` makes sense and already...

Another use case for `raw_predict` is #28574 which needs access to logits.

We discussed it in the monthly meeting https://github.com/scikit-learn/administrative/blob/master/monthly_meetings/2025-02-24.md and agreed on having a `raw_predict`, officially. And yes, for non-probabilistic classifiers, it returns the same as `decision_function`. We also discussed deprecating...

@agramfort @rth You might be interested.

Meanwhile, https://github.com/Quantco/glum uses this trick with https://github.com/Quantco/tabmat.

Adoption of narwhals in scikit-learn is now properly discussed in https://github.com/scikit-learn/scikit-learn/issues/31049. I also opened a draft PR https://github.com/scikit-learn/scikit-learn/pull/31127 to experiment. Unfortunately, it reveals problems: I wanted to use narwhals in...

Not supporting tuples is fine. But ```python df_nw[:, [0, 1]] # works df_nw[:, np.array([0, 1])] # raises error ``` For polars, both works fine, pandas also errors (in scikit-learn, this...

@jbogaardt Do you have a rough idea where this bug origins? What I could contribute, if it helps, is to add the above snippet as a test.

> Nothing around quantiles has anything to do with probabilities That's just not true. Ask any a statistician or refer to literature. Quantiles are, in a way, the inverse of...

> We're discussing the computational method here, not any statistical interpretation of it (which may or may not be relevant for the user). First, this issue is not about computational...