>The output of most machine learning algorithms is just a belief function. It's not even a belief function, in the sense of a normalized probability distributi

>The output of most machine learning algorithms is just a belief function. It's not even a belief function, in the sense of a normalized probability distribution that respects conditionalization properly. It's basically just a one-hot vector for classification.