Asymptotic Error Rates in Quantum Hypothesis Testing

TL;DR

This paper derives the asymptotic error rates for quantum hypothesis testing, generalizing classical distances and providing a unified, full generality proof with new techniques, highlighting the quantum Chernoff distance as a natural measure.

Contribution

It presents a unified, comprehensive proof of quantum asymptotic error rates, including the quantum Chernoff distance, for states with different supports, using new techniques.

Findings

Quantum Chernoff distance identified as a natural measure.

Unified proof for quantum hypothesis testing error rates.

Extension to states with different supports.

Abstract

We consider the problem of discriminating between two different states of a finite quantum system in the setting of large numbers of copies, and find a closed form expression for the asymptotic exponential rate at which the specified error probability tends to zero. This leads to the identification of the quantum generalisation of the classical Chernoff distance. The proof relies on two new techniques that have been introduced in [quant-ph/0610027] and [quant-ph/0607216], respectively, and that are also well suited to prove the quantum generalisation of the Hoeffding bound, which is a modification of the Chernoff distance and specifies the optimal achievable asymptotic error rate in the context of asymmetric hypothesis testing. This has been done subsequently by Hayashi [quant-ph/0611013] and Nagaoka [quant-ph/0611289] for the special case where both hypotheses have full support. Moreover, quantum Stein's Lemma and quantum Sanov's theorem may be derived directly from the quantum Hoeffding bound combining it with a result obtained recently in [math/0703772]. The goal of this paper is to present the proofs of the above mentioned results in a unified way and in full generality (allowing hypothetic states with different supports). Additionally, we give an in-depth treatment of the properties of the quantum Chernoff distance. We argue that, although it is not a metric, it is a natural distance measure on the set of density operators, due to its clear operational meaning.