Abstract:We prove the expected disturbance caused to a quantum system by a sequence of randomly ordered two-outcome projective measurements is upper bounded by the square root of the probability that at least one measurement in the sequence accepts. We call this bound the Gentle Random Measurement Lemma.
We then consider problems in which we are given sample access to an unknown state $\rho$ and asked to estimate properties of the accepting probabilities $\text{Tr}[M_i \rho]$ of a set of measurements $\{M_1, M_2, \ldots , M_m\}$. We call these types of problems Quantum Event Learning Problems. Using the gentle random measurement lemma, we show randomly ordering projective measurements solves the Quantum OR problem, answering an open question of Aaronson. We also give a Quantum OR protocol which works on non-projective measurements but which requires a more complicated type of measurement, which we call a Blended Measurement. Given additional guarantees on the set of measurements $\{M_1, \ldots, M_m\}$, we show the Quantum OR protocols developed in this paper can also be used to find a measurement $M_i$ such that $\text{Tr}[M_i \rho]$ is large. We also give a blended measurement based protocol for estimating the average accepting probability of a set of measurements on an unknown state.
Finally we consider the Threshold Search Problem described by O'Donnell and Bădescu. By building on our Quantum Event Finding result we show that randomly ordered (or blended) measurements can be used to solve this problem using $O(\log^2(m) / \epsilon^2)$ copies of $\rho$. Consequently, we obtain an algorithm for Shadow Tomography which requires $\tilde{O}(\log^2(m)\log(d)/\epsilon^4)$ samples, matching the current best known sample complexity. This algorithm does not require injected noise in the quantum measurements, but does require measurements to be made in a random order and so is no longer online.
What problem does this paper attempt to address?
This paper aims to solve the problem of perturbation caused by quantum measurement to quantum systems, especially how to quantify and control this perturbation in a series of two - outcome projection measurements in random order. Specifically, the authors focus on:
1. **Gentle Random Measurement Lemma**: It is proved that the expected perturbation of a sequence of two - outcome projection measurements in random order on a quantum system is upper - bounded within the square root of the probability that at least one measurement in this sequence is accepted. This result provides a new tool for understanding the impact of random measurement sequences on quantum systems.
2. **Quantum OR Problem**: Given an unknown quantum state \(\rho\) and a set of measurements \(\{M_1, M_2,\ldots, M_m\}\), the goal is to estimate the properties of the acceptance probabilities \(\text{Tr}[M_i\rho]\) of these measurements. In particular, the authors solve an open problem proposed by Aaronson, that is, whether projection measurements in random order can solve the Quantum OR problem. They prove that (with a small modification) Aaronson's original algorithm does work and can be used for other event - learning tasks, such as Shadow Tomography, thus simplifying the algorithms for these tasks.
3. **Quantum Event Finding**: This is a variant of the Quantum OR problem. It not only requires determining whether there is a measurement \(M_i\) such that \(\text{Tr}[M_i\rho]\) is large, but also needs to return it when such a measurement is found. The authors extend the Quantum OR algorithm to solve this problem and prove that under certain conditions, measurements in random order can effectively find such a measurement.
4. **Quantum Threshold Search**: The authors consider the threshold search problem described by O’Donnell and Bădescu. By constructing their Quantum Event Finding results, they show that random - order (or mixed) measurements can solve this problem using \(O(\log^2(m)/\epsilon^2)\) copies of \(\rho\). This leads to a Shadow Tomography algorithm that requires \(\tilde{O}(\log^2(m)\log(d)/\epsilon^4)\) samples, which matches the currently known best sample complexity.
5. **Quantum Mean Estimation**: Given a set of measurements \(\{M_1, M_2,\ldots, M_m\}\) and sample access to an unknown quantum state \(\rho\), the goal is to estimate the average acceptance probability \(\frac{1}{m}\sum_{i}\text{Tr}[M_i\rho]\) of these measurements. The authors propose a natural algorithm and analyze its performance, and find that the average value of non - commutative measurements can sometimes be estimated to a fixed precision with fewer samples than classical measurements.
### Main Results
1. **Gentle Random Measurement Lemma**:
- Theorem 1: Let \(M = \{M_1, M_2,\ldots, M_m\}\) be a set of two - outcome projection measurements, and \(\rho\) be a density matrix. Consider the process of uniformly randomly selecting a measurement from the set \(M\) and applying it to a quantum system with the initial state \(\rho(0)=\rho\). Let \(\rho(k)\) be the state of the quantum system after repeating this process \(k\) times without any measurement being accepted, then
\[
\|\rho - \rho(k)\|_1\leq4\sqrt{\text{Accept}(\lceil k/2\rceil)}\leq4\sqrt{\text{Accept}(k)}
\]