Discrimination of arbitrary unknown quantum channels and measurements

Marcin Markiewicz, Łukasz Pawela, Zbigniew Puchała · 2025

Read Discrimination of arbitrary unknown quantum channels and measurements by Marcin Markiewicz, Łukasz Pawela, Zbigniew Puchała (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17317.txt.

Introduction

Marcin Markiewicz, Łukasz Pawela, and Zbigniew Puchała analyze how well a single test can tell apart two randomly chosen quantum operations, in a paper from 2025. The operations may be channels, which transform quantum states, or measurements. Only one shot is allowed, so the protocol's success chance is the key measure.

The authors derive tight bounds on that success probability. The bounds hold for arbitrary channels and for generalized measurements, with no special assumptions about the pair drawn. Their efficiency analysis shows exactly how much a single-shot comparison can promise. The results give a firm limit for tasks that must judge quantum devices or procedures with minimal testing.

Catalog record

CREATORMarcin Markiewicz, Łukasz Pawela, Zbigniew Puchała
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-17317

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Discrimination of arbitrary unknown quantum channels and measurements Marcin Markiewicz, Łukasz Pawela, Zbigniew Puchała Abstract In this work we provide an efficiency analysis of the problem of discrimination of two randomly chosen unknown quantum operations in the single-shot regime. We provide tight bounds for the success probability of such a protocol for arbitrary quantum channels and generalized measurements. 1 Introduction and motivation Discrimination of unknown quantum operations is a task in which one is given an ability to apply two given operations with specified input and output dimensions on arbitrary quantum state, and decide on whether they are identical or distinct based on final measurement outcomes. Several classes of channels have been already analyzed in this context (unitary channels [ 1 , 2 , 3 ] , mixed unitary channels [ 4 ] , von Neumann measurements [ 5 , 6 , 7 ] , however there are no known results regarding discrimination of arbitrary channels and measurements in the most general discrimination scenario. Partial results on this issue are known only for the scenario of unambiguous discrimination of general measurements [ 8 , 9 ] , in which one assumes that the protocol in each run gives either perfectly correct answer or gives no answer at all. On the other hand in the scenario discussed in this work one assumes that the algorithm gives definite answer in each run, however the answer can be erroneous in some percentage of runs. Although any completely positive trace preserving (CPTP) channel can be seen from the dilation perspective as a unitary channel on an extended Hilbert space, the specific optimal performance of discrimination task significantly depends on the dimension of the extension and on the way of sampling the unknown channel, therefore it cannot be directly reduced to the unitary case itself. In this work we provide analysis of optimal performance of discrimination of unknown quantum channels and a related issue of unknown generalized quantum measurements in the single-shot scenario. The work is divided into two parts. Firstly we analyze the performance of discrimination of two random channels as a function of dimensions, which characterize it. In the second part we analyze discrimination of two random POVM-type measurements, based on results on discrimination of arbitrary channels. This can be done since a random POVM measurement, treated as a channel with classical output, can be represented as a composition of a random channel with the dephasing channel. Our results provide the most general solution to the discrimination problem of two quantum operations in the single-shot regime. 2 Discrimination of unknown channels 2.1 Symmetric discrimination Let us consider a scenario in which we are given two black boxes, each containing an unknown quantum channel, with specified input and output dimensions as respectively d_{i} and d_{o} . Our goal is to determine whether the boxes contain the same operation or a different ones, and provide a bound on the probability of correct guessing in this scenario. For a sketch see Fig. 1 . Figure 1: A schematic representation of the single-shot symmetric discrimination scheme. Note that the scheme allows for usage of an additional quantum register untouched by both channels. A bound on the probability of correct distinction between two channels \Phi and \Psi can be bounded by the diamond norm: p^{s}_{H}(\Phi,\Psi)\leq\frac{1}{2}+\frac{1}{4}\left\|\Phi-\Psi\right\|_{\diamond}, (1) which can be seen as an old Holevo-Helstrom result [ 10 , 11 ] restated using Kitaev’s diamond norm [ 12 ] . Let us denote the set of d\times d complex matrices by \mathcal{M}_{d}(\mathbb{C}) . For a Hermiticity-preserving \Xi:\mathcal{M}_{d_{i}}(\mathbb{C})\to\mathcal{M}_{d_{o}}(\mathbb{C}) with a Choi representation J(\Xi)\in\mathcal{M}_{d_{o}}(\mathbb{C})\otimes\mathcal{M}_{d_{i}}(\mathbb{C}) : J(\Xi)=\sum_{i,j=1}^{d_{i}}\Xi(\ket{i}\!\bra{j})\otimes\ket{i}\!\bra{j} we can provide an upper bound on the diamond norm [ 13 ] : \left\|\Xi\right\|_{\diamond}\leq\left\|\Tr_{d_{o}}\left|J(\Xi)\right|\right\|. (2) Since our task is to find performance of discrimination between unknown channels, we have to define a way of sampling such channels. To start, let us define an arbitrary channel \Phi_{U}^{(s)}:\mathcal{M}_{d_{i}}(\mathbb{C})\to\mathcal{M}_{d_{o}}(\mathbb{C}) via the Stinespring dilation with extension denoted as s : \Phi_{U}^{(s)}(\rho)=\Tr_{s}U\rho U^{\dagger}=\Tr_{s}\Phi_{U}(\rho), (3) where the input state \rho\in\mathcal{M}_{d_{i}}(\mathbb{C}) , U is an isometry operation such that: U:\mathbb{C}^{d_{i}}\to\mathbb{C}^{d_{o}}\otimes\mathbb{C}^{s}, (4) and \Phi_{U}(\rho)=U\rho U^{\dagger} denotes the corresponding isometry channel. Hence, a random quantum channel \Phi_{U}^{(s)} defined in this manner can be seen as a composition of a random isometry with a partial trace: \Phi_{U}^{(s)}=\Tr_{s}\circ\,\Phi_{U}. (5) Now we can define the measure \mu_{\Phi_{U}^{(s)}} on the set of quantum channels to be derived from to the Haar measure \mu^{s}_{d_{i}d_{o}} on the set of random isometries ( 4 ), which can be obtained as a measure on random truncated unitaries \textrm{U}(sd_{o}) [ 14 ] . Measure \mu^{s}_{d_{i}d_{o}} is a function of three dimensions: d_{i} , d_{o} and s which define the isometry ( 4 ). For the sake of clarity of further derivations we will denote the volume differential \differential\mu^{s}_{d_{i}d_{o}} corresponding to the measure \mu^{s}_{d_{i}d_{o}} as simply \differential U . When considering a quantum channel \Phi via Stinespring representation one needs to assume sd_{o}\geq d_{i} for the channel to be trace preserving. At the same time with the same assumption but different interpretation of s one can define two another measures on the set of quantum channels [ 15 ] , via random Choi matrices and random Krauss operators. As shown in [ 15 ] all these three ways of sampling random quantum channels are equivalent, in a sense that they define the same measure on the set of channels. For convenience in this work we solely utilise the Stinespring version. In the symmetric discrimination scenario we are discriminating between application of random channels \Phi_{U}^{(s)}\!\!\otimes\!\Phi_{U}^{(s)} (identical ones) and \Phi_{U}^{(s)}\!\!\otimes\!\Phi_{V}^{(s)} (distinct ones), drawn according to the measure \mu^{s}_{d_{i}d_{o}} . Such averaged channels are represented by the integrals \int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{U}^{(s)}\differential U and \int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{V}^{(s)}\differential U\differential V respectively. Hence, in order to find the Holevo-Helstrom bound ( 1 ) for the probability p^{s}_{H}\left(\int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{U}^{(s)}\differential U,\int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{V}^{(s)}\differential U\differential V\right) , we need to find the value of the following expression: \left\|\int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{U}^{(s)}\differential U-\int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{V}^{(s)}\differential U\differential V\right\|_{\diamond}. (6) In order to provide the diamond norm bounds we need the Choi representation of both the channels in ( 6 ). For now, we will focus on the first channel. As the only random part of the channel \Phi_{U}^{(s)} ( 5 ) is the isometry \Phi_{U} , from [ 7 , 16 ] we have: \begin{split}\mathcal{J}\left(\int\Phi_{U}\otimes\Phi_{U}\differential U\right)&=\frac{1}{(d_{o}s)^{2}-1}\left({\rm 1\kern-2.20001ptl}_{(d_{o}s)^{2}}\otimes{\rm 1\kern-2.20001ptl}_{d_{i}^{2}}+S_{(d_{o}s),(d_{o}s)}\otimes S_{d_{i},d_{i}}\right)+\\ &-\frac{1}{(d_{o}s)\left((d_{o}s)^{2}-1\right)}\left({\rm 1\kern-2.20001ptl}_{(d_{o}s)^{2}}\otimes S_{d_{i},d_{i}}+S_{(d_{o}s),(d_{o}s)}\otimes{\rm 1\kern-2.20001ptl}_{d_{i}^{2}}\right),\end{split} (7) where S_{a,b} denotes a swap operation between a b -dimensional system and an a -dimensional system. Composing this Choi matrix with the Choi representation of the partial trace via the notion of link product [ 17 ] we obtain a simple relation: \mathcal{J}\left(\Tr_{s}\circ\,\Phi_{U}\right)=\Tr_{s}\mathcal{J}(\Phi_{U}), (8) which gives, after straightforward yet tedious computations: \begin{split}\mathcal{M}_{d_{o}^{2}d_{i}^{2}}(\mathbb{C})\ni\mathcal{J}\left(\int\Phi_{U}^{(s)}\!\!\otimes\!\Phi_{U}^{(s)}\differential U\right)&=\frac{1}{(d_{o}s)^{2}-1}\left(s^{2}{\rm 1\kern-2.20001ptl}_{d_{o}^{2}}\otimes{\rm 1\kern-2.20001ptl}_{d_{i}^{2}}+sS_{d_{o},d_{o}}\otimes S_{d_{i},d_{i}}\right)+\\ &-\frac{1}{(d_{o}s)\left((d_{o}s)^{2}-1\right)}\left(s^{2}{\rm 1\kern-2.20001ptl}_{d_{o}^{2}}\otimes S_{d_{i},d_{i}}+sS_{d_{o},d_{o}}\otimes{\rm 1\kern-2.20001ptl}_{d_{i}^{2}}\right).\end{split} (9) As for the channel \int\Phi_{U}^{(s)}\!\otimes\Phi_{V}^{(s)}\differential U\differential V we follow the same reasoning as in [ 7 ] : \mathcal{J}\left(\int\Phi_{U}^{(s)}\!\otimes\Phi_{V}^{(s)}\differential U\differential V\right)=\frac{1}{d_{o}^{2}}{\rm 1\kern-2.20001ptl}_{d_{o}^{2}d_{i}^{2}}. (10) Finally, in order to apply the bound ( 2 ), we need to evaluate the absolute value operator of J\equiv\mathcal{J}\left(\int\Phi_{U}^{(s)}\!\otimes\Phi_{U}^{(s)}\differential U\right)-\mathcal{J}\left(\int\Phi_{U}^{(s)}\!\otimes\Phi_{V}^{(s)}\differential U\differential V\right) Consider a matrix W=S_{d_{o},d_{o}}\otimes S_{d_{i},d_{i}} . As shown in Appendix A we can construct a polar decomposition of the matrix J such that we have WJ=|J| . Hence we have: \displaystyle\Tr_{d_{o}}WJ \displaystyle= \displaystyle\Tr_{d_{o}}|J| \displaystyle= \displaystyle\frac{s(d_{o}^{2}-1)}{d_{o}^{2}s^{2}-1}{\rm 1\kern-2.20001ptl}_{d_{i}^{2}}-\frac{d_{o}^{2}-1}{d_{o}\left(d_{o}^{2}s^{2}-1\right)}S_{d_{i},d_{i}} \displaystyle= \displaystyle\frac{(d_{o}^{2}-1)}{d_{o}(s^{2}d_{o}^{2}-1)}\left(sd_{o}{\rm 1\kern-2.20001ptl}_{d_{i}^{2}}-S_{d_{i},d_{i}}\right). Using the fact that \|\alpha{\rm 1\kern-2.20001ptl}-S\|=1+\alpha,\;\alpha>0 , we finally obtain: \left\|\Tr_{d_{o}}\left|J\right|\right\|=\frac{d_{o}^{2}-1}{d_{o}^{2}s-d_{o}} (11) Hence, we get the

Read the complete text