Physics-Constrained Compressed Sensing for Quantum Sensing in the Data-Starved Regime

Amir Kalev

REVIEW 3 major objections 4 minor 31 references

Physics-Constrained Compressed Sensing for Quantum Sensing in the Data-Starved Regime

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Enforcing quantum-mechanical structure on correlation data lets quantum sensors recover frequencies from sparse, noisy samples.

desk verdict The empirical method looks useful in sparse-data quantum sensing, but the analytic claims rest on a false uniqueness lemma and a trace-minimization objective that is constant for Toeplitz matrices. read the letter →

arxiv 2608.11092 v1 pith:3NPMOOVJ submitted 2026-08-11 quant-ph

Amir Kalev This is my paper · ORCID

classification quant-ph

keywords quantumsensingcompressedcorrelationfunctionGrammatrixpositivesemidefiniteToeplitzstructurefrequencyestimationdata-starvedregime

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that a universal fact of quantum mechanics—that two-time correlation functions of Hermitian observables form positive-semidefinite matrices whose entries depend only on the time difference—can be used as a denoising prior for quantum sensing. It formulates signal reconstruction as a convex optimization program that enforces these structural constraints together with a low-rank prior, and proves that the true signal is uniquely recoverable from $O(r \log K)$ samples in the noiseless case and stably recoverable in noise. In a GHZ-based magnetometry example, the method reduces frequency-estimation error by up to more than an order of magnitude compared to direct fitting or matrix pencil when only 8–16 noisy time samples are available. This is a classical post-processing step that requires no extra hardware or calibration, and it does not change the fundamental quantum limit, yet it recovers information that noise and sparse sampling would otherwise destroy.

What carries the argument

The load-bearing object is the Gram matrix of the sampled correlation function, $G_{ij}=G(t_i-t_j)$, which every physical signal must endow with two properties: positive semidefiniteness (no negative eigenvalues) and Toeplitz structure (entries constant along diagonals, reflecting stationarity). The method's engine is the convex program (7), which minimizes the trace of $G$—a proxy for rank that is exact for PSD matrices—subject to matching the observed noisy samples and to these structural constraints. Because the feasible set is convex, the problem can be solved by standard semidefinite programming in $O(K^3)$ time, independent of the number of qubits. A companion uniqueness argument, based on the Haynsworth inertia additivity formula, shows that a PSD matrix of rank $r$ is uniquely determined by any $r$ of its rows and columns; this is what makes recovery from few samples possible.

What would settle it

Perform the GHZ magnetometry benchmark with $n=10$ qubits and $K=12$ noisy samples as in Section III, and measure the frequency-estimation error averaged over many noise realizations; the central claim predicts that the PSD-constrained reconstruction consistently beats both direct fitting and the matrix-pencil method, with errors approaching $10^{-4}$ while the baselines stay near $10^{-2}$.

Watch

Extended reading notes

Core claim

The central claim is that the physically necessary structure of correlation functions is itself enough to rescue parameter estimation in the data-starved regime. Any signal $G(t)=\langle O(t)O(0)\rangle$ from a Hermitian observable under unitary dynamics gives rise to a Gram matrix $G$ that is positive semidefinite and Toeplitz; when the dynamics is effectively low-rank (a handful of frequencies), $G$ has small rank. The paper's convex program (7)—minimize the trace of $G$ subject to the data-fitting constraint, the Toeplitz constraint, and $G \succeq 0$—is shown in Propositions 1 and 2 to recover the exact ground-truth matrix from $O(r \log K)$ observed samples in the noiseless case and to be stable under bounded noise, with reconstruction error proportional to the noise level. In the GHZ magnetometry benchmark, this reconstruction yields frequency estimates whose error is consistently lower than direct fitting of noisy samples and the matrix-pencil method throughout the data-starved regime, with improvements ranging from a factor of a few to more than an order of magnitude. The method does not augment the quantum Fisher information; it is a classical inference improvement that exploits the geometry of physical signals to undo the damage of finite sampling and imperfection noise.

Load-bearing premise

The method assumes that the noiseless correlation function is exactly a stationary sum of a few frequencies, so that its Gram matrix has low rank; if the true signal contains a broad spectral continuum or nonstationary correlations, the model is misspecified and the recovery guarantee does not apply.

Editorial extensions

If this is right

Quantum sensing protocols that measure time-domain correlation functions can gain improved precision from noisy, sparse data through classical post-processing alone, with no added hardware or calibration.
The method applies to any Hermitian observable under unitary dynamics, not just the GHZ toy model; the PSD constraint is universal, so the same reconstruction program transfers to Ramsey interferometry, NV-center magnetometry, and dynamical-decoupling spectroscopy.
In the data-starved regime (between 8 and 16 samples), the PSD-constrained reconstruction reduces frequency-estimation error by up to more than an order of magnitude relative to direct fitting and matrix pencil.
Under suitable sampling, the reconstruction error is bounded by a constant times the noise level, and the sample complexity scales as $O(r \log K)$, where $r$ is the number of frequency components.
Even under hidden dephasing, a model mismatch the estimator is not told about, the PSD-constrained reconstruction remains more accurate than unconstrained baselines, though all methods degrade as the mismatch grows.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same PSD-Toeplitz constraint could be applied directly to noise-spectroscopy data, such as dynamical-decoupling measurements, to denoise the inferred spectral density without assuming a specific noise model.
Because the reconstruction is a convex projection onto a physical set, it could be combined with adaptive experiment design: after an initial sparse sample, the reconstructed signal could guide which time points to measure next, potentially reducing total measurement time.
The rank-$r$ assumption is a strong implicit claim that the signal is a sum of a few undamped oscillators; a broader test with broadened spectral lines (Lorentzian or $1/f$ spectra) would map the actual operating range of the method.
The $O(r \log K)$ sampling guarantee suggests a practical rule of thumb: for a sensor with $r$ dominant frequencies, measuring about $r \log K$ time points and running the projection could approach dense-sampling performance at a fraction of the experimental cost.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Dear Colleague,

The useful part of this paper is the numerical demonstration: enforcing PSD and Toeplitz structure on correlation functions from GHZ magnetometry gives frequency estimates that beat direct fitting and matrix pencil when the number of time samples is small. The authors are honest about where the advantage shrinks, and the code is linked. That part is worth a look.

The trouble is the theory. Appendix A claims that the first r rows and columns of a rank-r PSD matrix determine the whole matrix within the PSD cone. That is false. For a Toeplitz example, take G0=1, G1=0.5; both G2=0 and G2=0.25 give valid 3x3 PSD Toeplitz matrices. So Propositions 1 and 2 are not established. The import from Ref. [25] is also not on point: that paper deals with strictly-complete measurements chosen for tomography, not fixed Toeplitz sampling.

There is a second, more basic issue. For a Toeplitz matrix, the trace is K times G0. If G0 is among the measured samples, the objective in program (7) is constant over the entire feasible set. So the 'trace minimization' is vacuous; the solver is just finding an arbitrary feasible point. That kills the low-rank-promotion narrative entirely. The empirical success then has to be attributed to the PSD+Toeplitz constraints alone, which do not give unique recovery, as the counterexample shows.

I want to be fair: the numerical results may still be real. The hidden-dephasing test and the operating-regime scan are the kind of honest checks I like to see. But the paper's headline claim of analytic guarantees is unsupported, and the current write-up gives a false proof in the appendix. That is load-bearing, not a cosmetic issue.

Who should read this? People building classical post-processing for quantum sensors in the low-sampling regime will find the empirical pattern intriguing and worth testing. But I would not cite the theoretical guarantees.

Recommendation: send to peer review with a request for major revision. Ask the authors to either prove a genuine recovery guarantee for the Toeplitz/PSD setting (with the trace issue addressed) or explicitly reframe the paper as an empirical study of a heuristic with a mapped operating regime. The code should be pinned to a commit. On the current math, the propositions should not stand as stated.

Referee Report

3 major / 4 minor

Summary. The manuscript proposes a physics-constrained compressed sensing framework for reconstructing two-time correlation functions in quantum sensing. The method enforces positive semidefiniteness (PSD) and Toeplitz structure on the correlation-function Gram matrix via the convex program (7), with trace minimization as a low-rank proxy. The paper claims analytic guarantees (Propositions 1 and 2) for exact recovery and stability, and reports numerical demonstrations on GHZ-based magnetometry showing improved frequency estimation in the data-starved regime, with code available at [23].

Significance. If the analytic guarantees were valid, the paper would provide a calibration-free, hardware-independent improvement for a broad class of quantum sensing protocols, and the numerical results suggest a practical benefit in sparse-sampling regimes. The manuscript also includes useful robustness tests (Appendix B.4, hidden dephasing) and a public code repository, which are strengths. However, the analytic claims are not established: the key uniqueness assertion in Appendix A is false, and the imported results from Refs. [19,25] do not apply to the Toeplitz sampling pattern. The numerical evidence may be sufficient to support a weaker, purely empirical claim, but the current central claim of analytic guarantees is unsupported.

major comments (3)

[Appendix A.2, Eqs. (A7)-(A9); Propositions 1-2] The uniqueness lemma is false. The paper asserts that the first r rows and columns of a rank-r PSD matrix uniquely determine it within the entire PSD cone. This is contradicted by a simple Toeplitz example: for a 3x3 real Toeplitz PSD matrix with G0=1 and G1=0.5, both G2=0 and G2=0.25 yield valid PSD Toeplitz matrices, so the data with r=1 do not determine G*. More generally, in the block form [[A,B],[B†,C]], replacing C by C+D with D⪰0 preserves PSD and the known submatrices A and B. The Toeplitz constraint does not remove this freedom in general. Consequently, the derivations of Proposition 1 (exact recovery) and Proposition 2 (stability) collapse; those results are not proven by the arguments in the appendix.
[Appendix A.2, paragraph 'With this structure of G*, we can directly import...'] The import of results from Refs. [19,25] is not justified. Ref. [25] concerns strictly-complete measurements for bounded-rank quantum-state tomography, a condition on the measurement operator that the fixed Toeplitz sampling pattern is not shown to satisfy. Ref. [19] establishes compressed sensing guarantees for random measurement maps satisfying the restricted isometry property, not for deterministic consecutive time samples. Since neither condition is verified for the sampling scheme in Eq. (7), the statement that the results 'directly import' is unsupported.
[Proposition 2 and Appendix A.2, Eq. (A9)] Even accepting the uniqueness claim, the stability bound in Eq. (8) is not derived. The appendix's program (A9) uses only the first r samples, whereas the main program (7) uses an arbitrary observed set Ω of size O(r log K). The paper does not show how a guarantee for the former transfers to the latter, nor does it provide the constant C in Eq. (8). The sentence 'it was proven in Ref. [25] that in the presence of noise...' is an assertion, not a proof, and it is not applicable to the present sampling model.

minor comments (4)

[Section II, after Eq. (2)] There is a typographical error: 'this follows form' should be 'this follows from'.
[Eqs. (3) and (7)] The notation Toeplitz({G_k}) is not precisely defined. Please specify whether G is Hermitian or real Toeplitz, and explicitly give the relation between the sequence {G_k} and the matrix entries (e.g., G_{ij} = G_{|i-j|} or G_{j-i}).
[Section III, Figs. 1 and 2] The figures are referenced but not reproduced in the manuscript text, making it difficult to assess the claimed numerical improvements. Please include the figures or specify where they can be found.
[Appendix B.2, Fig. 5] The robustness of results to the SDP tolerance ϵ is presented qualitatively. Reporting numerical values or a table for the estimation variance across ϵ would strengthen the claim that the choice ϵ=0.25 is not sensitive.

Circularity Check

2 steps flagged · score 8.0 of 10

Analytic recovery guarantees are imported wholesale from the author's own prior theorem rather than derived; the numerical tests are independent and keep the paper from maximal circularity.

uniqueness imported from authors [Appendix A.2, Application to Gram matrix recovery; basis for Propositions 1–2 in Section II]
"In Ref. [25] it was shown that if ρ is PSD matrix of rank r, then its first r rows and columns (or any of its r rows and corresponding columns, for that matter) uniquely specify it within the set of PSD matrices of dimension d, regardless of their rank, under the stated assumptions. ... With this structure of G⋆, we can directly import the results form Ref. [25] to the case of quantum sensing."

The paper's uniqueness guarantee is not derived here; it is lifted from Ref. [25], whose authors include the present author (Baldwin, Deutsch, and Kalev). That imported statement is exactly the premise needed to conclude that the first r correlation samples determine the full Toeplitz Gram matrix and hence that (7)/(A9) recovers G⋆. Because the theorem is cited rather than re-proved, and is applied to a Toeplitz top-row sampling model outside the strictly-complete measurement setting of Ref. [25], the analytic support for Propositions 1–2 is a self-citation chain.
self citation load bearing [Section II, Propositions 1–2, with proof deferred to Appendix A]
"Building on techniques from low-rank matrix recovery [19, 20] and the structured setting of Ref. [14], one can show results of the following form (see Appendix A): Proposition 1 (Exact recovery, noiseless case). Let G⋆ ∈ C be a rank-r Gram matrix. Under suitable sampling conditions on Ω, there exists |Ω|=O(r log K) (up to constants depending on the sampling geometry) such that G⋆ is the unique minimizer of Eq. (7) with ϵ=0."

The advertised analytic results are phrased as 'one can show' and deferred to Appendix A; Appendix A then says 'we can directly import the results form Ref. [25]' instead of proving them for Toeplitz Gram matrices. Refs. [19] and [25] are coauthored by the present author, so the theoretical content of Propositions 1–2 reduces to those prior papers. This is not an instance of independent, machine-checked or externally verified support: the cited theorem is the same uniqueness conclusion the paper needs, and no derivation specific to the present Toeplitz setting is supplied.

full rationale

The numerical claims are self-contained: the simulations, baselines, and code repository provide independent evidence that PSD/Toeplitz projection can improve frequency estimation from sparse noisy data. That portion is not circular. The circularity concerns the paper's analytic headline: 'We show analytically that ... the ground-truth signal can be uniquely identified in the noiseless case and recovered stably in the presence of noise.' The only derivation offered is Appendix A, which imports the uniqueness and stability statements from Ref. [25], a paper coauthored by the present author, and from Ref. [19], also coauthored by the present author. The imported statement—first r rows/columns of a rank-r PSD matrix determine it among all PSD matrices—is exactly the conclusion needed for Propositions 1–2, so the 'prediction' of exact/stabilized recovery is the prior result restated rather than a derived consequence. The self-citation is load-bearing because without it no analytic guarantee remains; the appendix contains no independent proof, and the cited theorem is not machine-checked or re-derived for Toeplitz data. Simple completions (e.g., PSD blocks differing only in the lower-right block) show the imported uniqueness statement is not automatic in this setting, reinforcing that the citation is doing the work. A score of 8 reflects that the theoretical result is forced by the self-citation chain, while the independent numerics prevent a 10.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the PSD/Toeplitz structural assumptions borrowed from Ref. [14], the low-rank prior, and an imported matrix-completion theorem from the author's prior work [25]. No new physical entities are introduced. The main free parameter is the SDP tolerance epsilon, which is user-set and shown to be non-critical.

free parameters (1)

epsilon (SDP tolerance) = 0.25
SDP tolerance chosen to ensure feasibility of constraint (7) given the simulated noise level. The paper reports that moderate variation changes the variance negligibly (App. B.2), so it is not a sharp fit, but it is a user-set constant in the reconstruction.

assumptions (5)

domain assumption Two-time correlation functions of Hermitian observables generate PSD Gram matrices (G >= 0).
Follows from the inner-product structure of quantum mechanics; introduced in Section II Eq. (2) and attributed to Ref. [14].
domain assumption Stationarity: the correlation function depends only on time differences, so the Gram matrix is Toeplitz.
Stated in Section II after Eq. (2). Applies to equilibrium or translationally invariant settings; not universal for all sensing protocols.
domain assumption The true correlation function has an effective low-rank spectral decomposition with r << K.
Assumed in Eq. (4) and used for the compressed-sensing advantage. The GHZ cosine signal is rank 2; broader spectra would violate this.
domain assumption Bounded l2 noise with known epsilon.
Eq. (6). Recovery guarantee Proposition 2 and the SDP constraint depend on the noise being bounded in l2 norm with a known tolerance.
standard math The matrix-completion uniqueness results of Ref. [25] apply to the Toeplitz setting.
Appendix A imports this result. As stated in the paper it is false among all PSD ranks; it holds for bounded rank or for the trace-minimizing solution. This is a load-bearing imported theorem.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Physics-Constrained Compressed Sensing for Quantum Sensing in the Data-Starved Regime." pith.science (2026). https://pith.science/paper/3NPMOOVJ

@misc{pith2026260811092,
  author       = {Pith},
  title        = {Pith review of: Physics-Constrained Compressed Sensing for Quantum Sensing in the Data-Starved Regime},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3NPMOOVJ}},
  note         = {Machine review of arXiv:2608.11092}
}

read the original abstract

Quantum sensors promise measurement sensitivities that can scale at the Heisenberg limit, but in practice their performance is often degraded by noise, finite sampling, and implementation imperfections. In this work we present a general framework for improving parameter estimation in such settings by exploiting intrinsic structural constraints of time-domain correlation functions. Our approach builds on the observation of Kemper et al. [PRL 132, 160403 (2024)] that two-time correlation functions of Hermitian observables generate Gram matrices that are positive semidefinite, a property that can be violated in experimentally acquired data. We formulate signal reconstruction as a convex optimization problem that enforces positive semidefiniteness, Toeplitz structure, and low-rank priors motivated by the underlying dynamics. We show analytically that, under suitable conditions, the ground-truth signal can be uniquely identified in the noiseless case and recovered stably in the presence of noise. We further demonstrate numerically, in a GHZ-based magnetometry protocol, that enforcing these physical constraints can significantly improve frequency estimation from sparse and noisy data. In particular, we observe a clear advantage in the data-starved regime, where only a small number of time samples are available and standard spectral estimation methods, including matrix pencil techniques, provide limited or unstable improvement over direct fitting. While the reconstructed signals do not in general reach the shot-noise-limited performance, the proposed approach consistently reduces estimation error and recovers much of the underlying structure of the signal. These results indicate that incorporating universal physical constraints into data analysis can enhance the practical performance of quantum sensing protocols without requiring additional hardware resources or calibration.

Figures

Figures reproduced from arXiv: 2608.11092 by the authors.

**Figure 2.** FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 5.** Figure 5: shows the dependence of the estimation variance on the numerical parameter ϵ. The parameter ϵ effectively determines how close the PSD solution is to the measured data, and it is set relative to the noise variance. In our simulations in the main text we used ϵ = 0.25…

**Figure 4.** Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

**Figure 6.** Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 26 canonical work pages

[25]

Strictly-complete measurements for bounded- rank quantum-state tomography,

Charles H. Baldwin, Ivan H. Deutsch, and Amir Kalev, “Strictly-complete measurements for bounded- rank quantum-state tomography,” Phys. Rev. A93, 052105 (2016). Appendix A: Compressed Sensing of a PSD Matrix

work page 2016
[23]

Amir Kalev,https://github.com/a-kalev/CS4QS (2025)

work page 2025
[19]

Quantum tomography protocols with positivity are compressed sensing protocols,

Amir Kalev, Robert L Kosut, and Ivan H Deutsch, “Quantum tomography protocols with positivity are compressed sensing protocols,” npj Quantum Informa- tion1, 1–6 (2015)

work page 2015
[1]

Quantum-enhanced measurements: Beating the stan- dard quantum limit,

Vittorio Giovannetti, Seth Lloyd, and Lorenzo Maccone, “Quantum-enhanced measurements: Beating the stan- dard quantum limit,” Science306, 1330–1336 (2004)

work page 2004
[2]

Neural-network heuristics for adaptive bayesian quan- tum estimation,

Lukas J. Fiderer, Jonas Schuff, and Daniel Braun, “Neural-network heuristics for adaptive bayesian quan- tum estimation,” PRX Quantum2, 020303 (2021)

work page 2021
[3]

Quantum metrology,

Vittorio Giovannetti, Seth Lloyd, and Lorenzo Mac- cone, “Quantum metrology,” Phys. Rev. Lett.96, 010401 (2006)

2006
[4]

A quantum rosetta stone for interferometry,

Hwang Lee, Pieter Kok, and Jonathan P. Dowling, “A quantum rosetta stone for interferometry,” Journal of Modern Optics49, 2325–2338 (2002)

work page 2002
[5]

High-sensitivity diamond magnetome- ter with nanoscale resolution,

J. M. Taylor, P. Cappellaro, L. Childress, L. Jiang, D. Budker, P. R. Hemmer, A. Yacoby, R. Walsworth, and M. D. Lukin, “High-sensitivity diamond magnetome- ter with nanoscale resolution,” Nature Physics4, 810–816 (2008)

work page 2008

Show all 31 references

[6]

Quantum sensing,

C. L. Degen, F. Reinhard, and P. Cappellaro, “Quantum sensing,” Rev. Mod. Phys.89, 035002 (2017)

2017
[7]

Toward heisenberg-limited spectroscopy with multiparti- cle entangled states,

D. Leibfried, M. D. Barrett, T. Schaetz, J. Brit- ton, J. Chiaverini, W. M. Itano, J. D. Jost, C. Langer, and D. J. Wineland, “Toward heisenberg-limited spectroscopy with multiparti- cle entangled states,” Science304, 1476–1478 (2004), https://www.science.org/doi/pdf/10.1126/s...

2004 doi
[8]

Improvement of frequency standards with quantum entanglement,

S. F. Huelga, C. Macchiavello, T. Pellizzari, A. K. Ekert, M. B. Plenio, and J. I. Cirac, “Improvement of frequency standards with quantum entanglement,” Phys. Rev. Lett. 79, 3865–3868 (1997)

1997
[9]

Adaptive quantum metrology under gen- eral markovian noise,

Rafa l Demkowicz-Dobrza´ nski, Jan Czajkowski, and Pavel Sekatski, “Adaptive quantum metrology under gen- eral markovian noise,” Phys. Rev. X7, 041009 (2017)

2017
[10]

End-to-end variational quantum sensing,

Benjamin MacLellan, Piotr Roztocki, Stefanie Czischek, and Roger G. Melko, “End-to-end variational quantum sensing,” npj Quantum Information10, 118 (2024)

2024
[11]

Bayesian quantum noise spec- troscopy,

Chris Ferrie, Chris Granade, Gerardo Paz-Silva, and Howard M. Wiseman, “Bayesian quantum noise spec- troscopy,” New Journal of Physics20, 123005 (2018)

2018
[12]

Qubits as spec- trometers of dephasing noise,

Kevin C. Young and K. Birgitta Whaley, “Qubits as spec- trometers of dephasing noise,” Phys. Rev. A86, 012314 (2012)

2012
[13]

Qubit noise spectroscopy for non-gaussian dephas- ing environments,

Leigh M. Norris, Gerardo A. Paz-Silva, and Lorenza Vi- ola, “Qubit noise spectroscopy for non-gaussian dephas- ing environments,” Phys. Rev. Lett.116, 150503 (2016)

2016
[14]

Denoising and extension of response functions in the time domain,

Alexander F. Kemper, Chao Yang, and Emanuel Gull, “Denoising and extension of response functions in the time domain,” Phys. Rev. Lett.132, 160403 (2024)

2024
[15]

Sensitivity optimization for nv- diamond magnetometry,

John F. Barry, Jennifer M. Schloss, Erik Bauch, Matthew J. Turner, Connor A. Hart, Linh M. Pham, and Ronald L. Walsworth, “Sensitivity optimization for nv- diamond magnetometry,” Rev. Mod. Phys.92, 015004 (2020). 7

2020
[16]

Measuring the spectrum of colored noise by dynamical decoupling,

Gonzalo A. ´Alvarez and Dieter Suter, “Measuring the spectrum of colored noise by dynamical decoupling,” Phys. Rev. Lett.107, 230501 (2011)

2011
[17]

Adaptive bayesian quantum tomography,

F. Husz´ ar and N. M. T. Houlsby, “Adaptive bayesian quantum tomography,” Phys. Rev. A85, 052120 (2012)

2012
[18]

Control-enhanced non-markovian quantum metrology,

Xiaodong Yang, Xinyue Long, Ran Liu, Kai Tang, Yue Zhai, Xinfang Nie, Tao Xin, Jun Li, and Dawei Lu, “Control-enhanced non-markovian quantum metrology,” Communications Physics7, 282 (2024)

2024
[20]

Quantum state tomography via compressed sensing,

David Gross, Yi-Kai Liu, Steven T. Flammia, Stephen Becker, and Jens Eisert, “Quantum state tomography via compressed sensing,” Phys. Rev. Lett.105, 150401 (2010)

2010
[21]

Michael A Nielsen and Isaac L Chuang,Quantum compu- tation and quantum information(Cambridge university press, 2010)

2010
[22]

Matrix pen- cil method for estimating parameters of exponentially damped/undamped sinusoids in noise,

Yingbo Hua and Tapan K. Sarkar, “Matrix pen- cil method for estimating parameters of exponentially damped/undamped sinusoids in noise,” IEEE Transac- tions on Acoustics, Speech, and Signal Processing38, 814–824 (1990)

1990
[24]

Tight oracle inequalities for low-rank matrix recovery from a minimal number of noisy random measurements,

E. J. Candes and Y. Plan, “Tight oracle inequalities for low-rank matrix recovery from a minimal number of noisy random measurements,” IEEE Trans. Inf. Theor. 57, 2342–2359 (2011)

2011
[26]

compressible

Background Compressed sensing (CS) is a signal processing tech- nique that allows for the efficient acquisition and recon- struction of a signal by exploiting its low-rankness (in the case where the signal is represented by a Hermitian ma- trix) [24]. Consider ad×dHermitian ma...
[27]

Application to Gram matrix recovery This subsection follows closely the methods presented in Ref. [25]. There, the authors extended the application of CS to PSD matrices where the acquisition map does not follow RIP (or other related properties). Instead, the acquired data cor...
[28]

alias- ing

Partial-period sampling In Fig. 4 we plot the varianceδαwhen the observations are limited to a fraction of the oscillation period. If the measurement covers only a fraction of an oscillation, the correlation function remains analytic and well-defined, and our PSD-based method,...
[29]

The parameterϵ effectively determines how close the PSD solution is to the measured data, and it is set relative to the noise vari- ance

Dependence on the SDP tolerance Figure 5 shows the dependence of the estimation vari- ance on the numerical parameterϵ. The parameterϵ effectively determines how close the PSD solution is to the measured data, and it is set relative to the noise vari- ance. In our simulations ...
[30]

In this test, we used K∈{6,8,10,12,14,16}, γ∈[0.05,0.3], with six equally spaced values ofγ

Operating regime in sample number and imperfection strength To further characterize the regime in which the PSD- constrained reconstruction is useful, we performed a two- dimensional numerical scan over the number of retained time samplesKand the imperfection-noise strengthγ. ...
[31]

In this test, the synthetic noiseless signal was replaced by GΓ(t)=e −Γt cos(nαt),(B1) where Γ is a dephasing rate

Robustness to hidden dephasing To address the robustness of the method to a more physically motivated noise mechanism, we also tested a dephasing model. In this test, the synthetic noiseless signal was replaced by GΓ(t)=e −Γt cos(nαt),(B1) where Γ is a dephasing rate. Importan...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [25]

Strictly-complete measurements for bounded- rank quantum-state tomography,

Charles H. Baldwin, Ivan H. Deutsch, and Amir Kalev, “Strictly-complete measurements for bounded- rank quantum-state tomography,” Phys. Rev. A93, 052105 (2016). Appendix A: Compressed Sensing of a PSD Matrix

work page 2016

[2] [23]

Amir Kalev,https://github.com/a-kalev/CS4QS (2025)

work page 2025

[3] [19]

Quantum tomography protocols with positivity are compressed sensing protocols,

Amir Kalev, Robert L Kosut, and Ivan H Deutsch, “Quantum tomography protocols with positivity are compressed sensing protocols,” npj Quantum Informa- tion1, 1–6 (2015)

work page 2015

[4] [1]

Quantum-enhanced measurements: Beating the stan- dard quantum limit,

Vittorio Giovannetti, Seth Lloyd, and Lorenzo Maccone, “Quantum-enhanced measurements: Beating the stan- dard quantum limit,” Science306, 1330–1336 (2004)

work page 2004

[5] [2]

Neural-network heuristics for adaptive bayesian quan- tum estimation,

Lukas J. Fiderer, Jonas Schuff, and Daniel Braun, “Neural-network heuristics for adaptive bayesian quan- tum estimation,” PRX Quantum2, 020303 (2021)

work page 2021

[6] [3]

Quantum metrology,

Vittorio Giovannetti, Seth Lloyd, and Lorenzo Mac- cone, “Quantum metrology,” Phys. Rev. Lett.96, 010401 (2006)

2006

[7] [4]

A quantum rosetta stone for interferometry,

Hwang Lee, Pieter Kok, and Jonathan P. Dowling, “A quantum rosetta stone for interferometry,” Journal of Modern Optics49, 2325–2338 (2002)

work page 2002

[8] [5]

High-sensitivity diamond magnetome- ter with nanoscale resolution,

J. M. Taylor, P. Cappellaro, L. Childress, L. Jiang, D. Budker, P. R. Hemmer, A. Yacoby, R. Walsworth, and M. D. Lukin, “High-sensitivity diamond magnetome- ter with nanoscale resolution,” Nature Physics4, 810–816 (2008)

work page 2008