The decision function can be seen as a distribution too. Say you're thinking of building a classifier, you can build the distribution over the domain x {-1,1}

The decision function can be seen as a distribution too. Say you're thinking of building a classifier, you can build the distribution over the domain x {-1,1} Give me f: A -> B and I'll give you P : (A x B) -> R+ See what I did there?