A discrimination problem consists of N linearly independent pure quantum states = {|φi} and the corresponding occurrence probabilities η = {ηi}. With any such problem, we associate, up to a permutation over the probabilities {ηi}, a unique pair of density matrices ρT and η p defined on the N-dimensional Hilbert space HN. The first one, ρT , provides a new parametrization of a generic full-rank density matrix in terms of the parameters of the discrimination problem, i.e., the mutual overlaps γi j = φi|φ j and the occurrence probabilities {ηi}. The second one, on the other hand, is defined as a diagonal density matrix η p with the diagonal entries given by the probabilities {ηi} with the ordering induced by the permutation p of the probabilities. ρT and η p capture information about the quantum and classical versions of the discrimination problem, respectively. In this sense, when the set can be discriminated unambiguously with probability one, i.e., when the states to be discriminated are mutually orthogonal and can be distinguished by a classical observer, then ρT → η p. Moreover, if the set lacks its independency and cannot be discriminated anymore, the distinguishability of the pair, measured by the fidelity F-ρT , η p-, becomes minimum. This enables one to associate with each discrimination problem a measure of discriminability defined by the fidelity F-ρT , η p-. This quantity, though distinct from the maximum probability of success, has the advantage of being easy to calculate, and in this respect, it can find useful applications in estimating the extent to which the set is discriminable.

Keywords