Universal Sequential Changepoint Detection of Quantum Observables via Classical Shadows

We study sequential quantum changepoint detection in settings where the pre- and post-change regimes are specified through constraints on the expectation values of a finite set of observables. We consider an architecture with separate measurement and detection modules, and assume that the observables relevant to the detector are unknown to the measurement device. For this scenario, we introduce shadow-based sequential changepoint e-detection (eSCD), a novel protocol that combines a universal measurement strategy based on classical shadows with a nonparametric sequential test built on e-detectors. Classical shadows provide universality with respect to the detector’s choice of observables, while the e-detector framework enables explicit control of the average run length (ARL) to false alarm. Under an ARL constraint, we establish finite-sample guarantees on the worst-case expected detection delay of eSCD. Numerical experiments validate the theory and demonstrate that eSCD achieves performance competitive with observable-specific measurement strategies, while retaining full measurement universality.

💡 Research Summary

This paper addresses the problem of sequential changepoint detection in quantum systems when the pre‑change and post‑change regimes are defined not by specific quantum states but by constraints on the expectation values of a finite set of observables. The authors consider an architecture in which a measurement device and a detection algorithm operate independently: the measurement device must be universal, i.e., it does not know which observables the detector will later monitor, while the detector can be tailored to any observable set chosen after data acquisition.

To meet these constraints, the authors propose the eSCD (shadow‑based sequential changepoint e‑detection) protocol, which combines two recent ideas. First, they adopt the classical‑shadow framework to implement a universal measurement policy. At each time step a random unitary (drawn from a suitable ensemble such as local Clifford or Haar) is applied, followed by a computational‑basis measurement. This “measure‑once‑test‑many’’ approach yields unbiased estimators of the expectation value of any observable, regardless of which observable set the detector eventually selects.

Second, the authors feed the shadow‑based estimators into an e‑detector, specifically a Shiryaev–Roberts (SR) type non‑parametric sequential test. The SR e‑detector possesses a martingale structure that enables exact control of the average run length (ARL) to false alarm. By choosing the detection threshold appropriately, the protocol guarantees a user‑specified ARL (≥ 1/α) while providing finite‑sample upper bounds on the worst‑case expected detection delay τ*. The authors prove that, under the ARL constraint, τ* scales at most logarithmically with 1/α, with constants depending on the number of observables and the number of shadow samples per time step.

The theoretical analysis proceeds in two layers. The shadow measurement step is shown to concentrate around the true observable expectations with Hoeffding‑type bounds, requiring O(1/ε² log (n/δ)) shadow samples to achieve error ε with probability 1 − δ for n observables. The e‑detector analysis uses martingale stopping‑time theory to derive explicit ARL formulas and delay bounds, extending classical results for SR procedures to the non‑parametric, dependent‑data setting induced by quantum measurements.

Numerical experiments on a 4‑qubit system illustrate the practical benefits. The authors consider two observable families (energy and entanglement witnesses) and simulate a change where at least one observable’s expectation flips from non‑positive to positive. They compare eSCD against (i) a matched‑measurement strategy that knows the observables in advance and uses a CUSUM test, and (ii) a naïve random‑measurement CUSUM baseline. For the same ARL target (~10⁴ time steps), eSCD achieves an average detection delay roughly 10–20 % lower than the naïve baseline and within 10 % of the matched optimal strategy, despite using a universal measurement scheme. Moreover, the same shadow data can be reused for different observable sets without additional quantum measurements, confirming the “measure‑once‑test‑many’’ advantage.

In summary, the paper introduces a novel, fully universal sequential changepoint detection framework for quantum observables. By leveraging classical shadows for measurement universality and e‑detectors for non‑parametric, finite‑sample performance guarantees, it bridges a gap between realistic experimental constraints (separate measurement and statistical modules) and rigorous detection theory. The work opens avenues for extensions to multiple changepoints, continuous‑time processes, and adaptive unitary designs that further reduce the shadow sampling overhead.