Sample-optimal learning of quantum states using gentle measurements

Cristina Butucea, Jan Johannes, Henning Stein · 2025

Read Sample-optimal learning of quantum states using gentle measurements by Cristina Butucea, Jan Johannes, Henning Stein (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-2505-24587.txt.

Introduction

Cristina Butucea, Jan Johannes, and Henning Stein present a 2025 paper on learning a quantum state with measurements that barely disturb it. A gentle measurement leaves the state close to its start while still giving data to learn from. The authors define a class of such measurements that act on each part of a product state on its own.

They prove a strong rule that limits how much information these measurements can pass, using a sharpened link between gentleness and privacy. From that rule, they derive the least number of copies needed to learn or certify a state to a set accuracy. They also give a concrete gentle method that reaches those bounds.

Catalog record

CREATORCristina Butucea, Jan Johannes, Henning Stein
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE106 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-24587

Browse the full library · About the Tower of Babel

Full text

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

Sample-optimal learning of quantum states using gentle measurements Cristina Butucea, Jan Johannes, Henning Stein Abstract Gentle measurements of quantum states do not entirely collapse the initial state. Instead, they provide a post-measurement state at a prescribed trace distance α \alpha from the initial state together with a random variable used for quantum learning of the initial state. We introduce here the class of α − \alpha- locally-gentle measurements ( α − \alpha- LGM) on a finite dimensional quantum system which are product measurements on product states and prove a strong quantum Data-Processing Inequality (qDPI) on this class using an improved relation between gentleness and quantum differential privacy. We further show a gentle quantum Neyman-Pearson lemma which implies that our qDPI is asymptotically optimal (for small α \alpha ). This inequality is employed to show that the necessary number of quantum states for prescribed accuracy ϵ \epsilon is of order 1 / ( ϵ 2 ​ α 2 ) 1/(\epsilon^{2}\alpha^{2}) for both quantum tomography and quantum state certification. Finally, we propose an α − \alpha- LGM called quantum Label Switch that attains these bounds. It is a general implementable method to turn any two-outcome measurement into an α − \alpha- LGM. 1 Introduction Quantum measurements, which are physical devices to extract information about the quantum state or system of states at hand, are inherently destructive. This destruction is even reflected in the widely used term collapse of a wave function. Quantum instruments are physical manipulations that can produce both statistical information (a random variable is observed) and a resulting collapsed quantum state. Recently, the pioneering work by [ 1 ] formalized a class of measurements called gentle which produce both statistical information and a resulting state which is at a prescribed distance from the measured state. We call the resulting state a "post-measurement state" rather than a "collapsed state" to highlight that it is indeed not far away from the initial state. While some principles of gentle (or weak) measurements were already known to researchers [ 11 , 16 , 30 ] the introduced formalization sparked a renewed interest in the problem. In recent times, gentle measurements have been studied in a variety of contexts such as quantum event learning [ 60 ] , Quantum Information Leakage [ 24 ] , Quantum encryption [ 26 ] , quantum machine learning [ 2 ] , tests for group isomorphisms [ 29 ] and even the foundations of quantum theory [ 56 ] . In each of these cases, the properties of gentle measurement prove essential for the tasks studied. The upside of gentleness is that after measuring a state gently the resulting post-measurement state is still of practical value. Rather than collapsing into a completely unrelated state the post-measurement state for a gentle measurement is by definition close to the original one. As such, it may be used again to store or to extract further information from it. These properties have already been observed in physical experiments [ 37 ] showing the physical feasibility of gentle measurement. For study of their statistical properties, especially the connection between gentleness and quantum Differential Privacy (qDP) established in [ 1 ] is proving fruitful for us in the development of new results in both fields. While [ 1 ] only consider gentleness on a whole system, that we now call global gentleness , we give a new definition of local gentleness and locally-gentle measurements . In contrast to globally-gentle measurements their local counterparts are unentangled and measure each subsystem of a quantum system one at a time. This approach allows for a far more physically feasible description of measurements [ 12 , 40 ] as holding and manipulating a large number of quantum states at one time surpasses current technological capabilities. With the ever rising capabilities of QPUs and accessibility to qubits we believe that it is important to study the properties of gentle measurements for those types of states. In a recent paper [ 2 ] discussed how gentle measurements may be necessary in order to implement quantum backpropagation algorithms for machine learning. Given the success of classical backpropagation algorithms, studying gentle measurements for qubits, which constitute the building blocks of any modern quantum computer, is a integral part in the study of advanced quantum machine learning algorithms. In this paper, we consider the problems of estimating and testing qubits using gentle measurements. We start by giving a brief introduction into the necessary concepts of quantum and classical statistics in Section 2 . In Section 3 we then introduce the concepts of gentle measurements. We state some basic properties of gentle measurements and improve the constant relating gentleness and differential privacy established by [ 1 ] and extend its applicability to a wider range of values for the gentleness parameter \alpha . Furthermore, we give a short result on how to construct gentle measurements. In Section 4 we then prove a proper quantum data-processing inequality (qDPI) for gentle measurements that relates the information extracted from a state and the destruction of that state by an arbitrary gentle measurement similar to a classical result for differentially-private mechanisms due to [ 20 ] . We demonstrate the power of our qDPI by deriving theoretical guarantees for both locally-gentle quantum state certification and locally-gentle quantum tomography in Section 5 . Moreover, in Section 6 , we develop a quantum locally-gentle measurement technique that we call the quantum Label Switch (qLS), implementable using well known tools of quantum computing, that allows for locally-gentle state certification and locally-gentle tomography for qubits. Given a gentleness parameter that the user can fix in advance, we build explicit procedures for both problems that use the qLS measurement and show that a loss of rate is unavoidable in both problems due to the gentleness constraints. Finally, some further proofs can be found in Section 8 . 2 Quantum statistics background Let us recall some basic definitions of both quantum and classical statistics. We will use the typical physicist notation where \ket{\psi}\in\mathcal{H} denotes a unit vector in a Hilbert space \mathcal{H} and \bra{\varphi}\in\mathcal{H}^{*}\cong\mathcal{H} its dual vector. Consequently, the inner product between \ket{\varphi} and \ket{\psi} can be written as \bra{\varphi}\ket{\psi} and the projection onto the space spanned by \ket{\psi} is denoted as \ket{\psi}\bra{\psi} . 2.1 States and measurements The laws of quantum mechanics dictate that a quantum system is described by a Hilbert space \mathcal{H} and that the state of the system is given by a unit vector \ket{\psi}\in\mathcal{H} . 1 1 1 We will not go into the exact origin of \ket{\psi} other than it being a solution to the famous ”Schrödinger equation” Even if we had complete knowledge of the state, any value we measure would still be probabilistic. Furthermore, when we have incomplete knowledge of the system, we must deal with the additional uncertainty of which state we are dealing with. From the outside, we are dealing with a probability distribution over a number of possible pure states \ket{\psi_i} . It turns out that the correct way to describe this probability distribution over pure states is as an operator \rho=\sum_{i}\lambda_{i}\ket{\psi_i}\bra{\psi_i} for orthogonal \ket{\psi_i} such that \sum_{i}\lambda_{i}=1 . This motivates the following definition. Definition 1 . A finite dimensional quantum system is described

Read the complete text