SVR inherited attributes: probB / probA

[ingrid88]

probA_ and probB_ are attributes inherited from the base class libsvm in the SVR class. However SVR does not allow for a probability = True as an input like the libsvm class does, therefore these attributes always return an empty array. I'm not sure they should be included here.

Link to code: https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/svm/classes.py#L824

[amueller]

They should indeed be removed / deprecated. I think someone might be working on that already (or did I talk to you about that earlier?)
See https://scikit-learn.org/dev/developers/contributing.html#deprecation

[amueller]

do you want to work on this?
You can talk to @meiguan who does something similar for #14766.

[ingrid88]

Ok sounds good

[algofrancesca]

We are working on the same issue but for OneClassSVM @eugeniaft

I think this is closed.

[amueller]

@Mzeej I don't think so, can you double check? There was no fixed merged, right?

Looks like it's still there on master:

from sklearn.datasets import make_regression
from sklearn.svm import SVR
X, y = make_regression()
svr = SVR().fit(X, y)
svr.probA_

[amueller]

@thomasjpfan can you remind me why these were closed without getting merged? The deprecation still seems reasonable and necessary.

[thomasjpfan]

@amueller They were closed by the contributors.