fitPosterior warning: Why does fitPosterior output a warning when the transformation is a step function?

I'm training various SVM models and saving the output from MATLAB. The output has lots of these warning messages:

Warning: Classes are perfectly separated. The optimal score-to-posterior transformation is a step function.

Is this not a good thing? Why does MATLAB output a warning? Is there a time this should concern someone?

Thanks.

Answer 1

[the cyclist]

I would say that a warning could be warranted because perfect separability could be an indication of over-fitting, and that the model may not generalize well to unseen data.