Statistical constructions in quantum information theory

TL;DR

This paper introduces statistical strategies in quantum nonlocal games, demonstrating their equivalence to existing quantum strategies and revealing new distinctions using recent mathematical results, with applications to the CHSH game.

Contribution

It defines statistical strategies and proves their equivalence to quantum strategies, using recent solutions to Tsirelson's problem to show new distinctions in strategy sets.

Findings

Statistical strategies are equivalent to quantum strategies for nonlocal games.

Existence of a nonlocal game where statistical commuting strategies exceed the closure of statistical spatial strategies.

Explicit statistical strategy for winning the CHSH game with probability above classical limits.

Abstract

We introduce a notion of strategies based on averaging for nonlocal games in quantum information theory. These so-called statistical strategies come in a commuting type and a more specific spatial type, which are respectively special cases of the quantum commuting and quantum spatial strategies commonly considered in the field. We prove a theorem that the sets of statistical commuting strategies and statistical spatial strategies are respectively equal to the sets of quantum commuting strategies and quantum spatial strategies for any nonlocal game. Thus we are able to use the recent negative solution of Tsirelson's problem to obtain a statistical analog showing that there exists a nonlocal game where the set of statistical commuting strategies properly contains the closure of the set of statistical spatial strategies. The proof of this theorem involves development of statistical replicas for numerous constructions in quantum information theory, in particular for the Fourier-type duality between observation structures and dynamical structures. The main point of the argument is to apply the established theory of approximating unitary representations of countable discrete groups by ergodic measure preserving actions of such groups. We note that the relevant groups are nonamenable. We also give an explicit description of a statistical strategy to win the CHSH game from Aspect's experiment with a probability exceeding the maximum possible value for a classical strategy.