On the role of designs in the data-driven approach to quantum statistical inference

Abstract

Designs, and in particular symmetric, informationally complete (SIC) structures, play an important role in the quantum tomographic reconstruction process and, by extension, in certain interpretations of quantum theory focusing on such a process. This fact is due to the symmetry of the reconstruction formula that designs lead to. However, it is also known that the same tomographic task, albeit with a less symmetric formula, can be accomplished by any informationally complete (non necessarily symmetric) structure. Here we show that, if the tomographic task is replaced by a data-driven inferential approach, the reconstruction, while possible with designs, cannot by accomplished anymore by an arbitrary informationally complete structure. Hence, we propose the data-driven inference as the arena in which the role of designs naturally emerges. Our inferential approach is based on a minimality principle according to which, among all the possible inferences consistent with the data, the weakest should be preferred, in the sense of majorization theory and statistical comparison.