DOC Add temperature scaling example #32854
Conversation
|
Given that we already have so many examples, see #27151, I‘d prefer to have this here integrated into an existing one. |
|
I'll convert this back to draft and work on it. |
virchan
left a comment
There was a problem hiding this comment.
Ready to review! Since plot_calibration_curve.py now covers multi-class classification, we can probably remove plot_calibration_multiclass.py.
@lorentzenchr, would you like to have a look?
|
Thanks for putting in an existing example. I am a bit undecided between the one you chose "Probability Calibration curves"/plot_calibration_curve.py and "Probability Calibration for 3-class classification"/plot_calibration_multiclass.py. |
|
I agree: i think we should keep this particular example about calibration of binary classifiers because a lot of the text in the analysis that was correct for binary classification (eg monotonicity and area under the roc curve) does not carry over directly to the multiclass setting. |
Reference Issues/PRs
A follow-up to #31068 and #32809 (comment)
What does this implement/fix? Explain your changes.
This PR adds an example demonstrating the use of temperature scaling for multi-class probability calibration.
Any other comments?
Cc @lorentzenchr for visibility