The Geometry of Losses

Loss functions are central to machine learning because they are the means by which the quality of a prediction is evaluated. Any loss that is not proper, or can not be transformed to be proper via a l