Which of the following transformations would make sense for transforming the raw model output of a linear classifier into a class probability?
