“…However, those approaches are characterized by a lack of generalisability, and the use of strong assumptions [7]. Imprecise probabilities [34,37,12,13], on the other hand, can provide a radically different solution to the construction of bounds in learning theory. A hierarchy of formalisms aimed at mathematically modeling the "epistemic" uncertainty induced by sources such as lack of data, missing data or data which is imprecise in nature [17,30,31], e.g.…”