When is an input state always better than the others ?: universally optimal input states for statistical inference of quantum channels

Abstract

Statistical estimation and test of unknown channels have attracted interest of many researchers. In optimizing the process of inference, an important step is optimization of the input state, which in general do depend on the kind of inference (estimation or test, etc.), on the error measure, and so on. But sometimes, there is a universally optimal input state, or an input state best for all the statistical inferences and for all the risk functions. In the paper, the existence of a universally optimal state is shown for group covariant/contravariant channels, unital qubit channels and some measurement families. To prove these results, theory of "comparison of state families" are used. We also discuss about effectiveness of entanglement and adaptation of input states.