pith. machine review for the scientific record. sign in

arxiv: 2604.12323 · v3 · submitted 2026-04-14 · 🪐 quant-ph

Recognition: unknown

Quantum-Enhanced Single-Parameter Phase Estimation with Adaptive NOON States

Simanshu Kumar , Nandan S Bisht
Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:07 UTC · model grok-4.3

classification 🪐 quant-ph
keywords quantum metrologyNOON statesphase estimationFisher informationgradient optimizationquantum sensingentangled photons
0
0 comments X

The pith

Optimizing eight circuit parameters via gradient descent on Fisher information raises NOON-state phase sensing from 36 percent to 58 percent of the Heisenberg limit at N=5 and multiplies useful events per pulse by up to 133 times.

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

The paper demonstrates that conventional NOON-state setups for single-parameter phase estimation leave substantial performance untapped. By implementing an end-to-end differentiable optical simulation and applying Adam gradient descent to maximize the classical Fisher information summed over coincidence channels, the authors identify circuit parameters that outperform the Afek et al. 2010 working point. Gains in raw information and post-selection efficiency increase sharply with photon number, turning quantum-enhanced sensing at N=3 and above into a realistic experimental prospect. The work supplies concrete numerical benchmarks showing how adaptive probe optimization can close much of the gap to the Heisenberg limit without requiring larger entangled states.

Core claim

Starting from numerically faithful reproductions of Afek coincidence fringes, Adam optimization of the eight trainable parameters in the hybrid coherent-plus-squeezed circuit increases classical Fisher information by 153 percent at N=2, 834-956 percent at N=3, 829-1598 percent at N=4, and 1775 percent at N=5. Post-selection rates rise by 153-3269 percent, yielding 8x to 133x more useful measurement events per pulse. Quantum Fisher information calculations confirm the optimized probe reaches 82 percent of the Heisenberg limit at N=2 and improves from 36 percent to 58 percent at N=5, with the inter-channel trade-off weakening at higher N.

What carries the argument

Differentiable maximization of the summed classical Fisher information across all coincidence channels by gradient descent on eight optical-circuit parameters in a Strawberry Fields simulation of adaptive NOON-state generation and measurement.

If this is right

  • Quantum-enhanced single-parameter phase estimation at N greater than or equal to 3 becomes experimentally practical because useful events per pulse increase by orders of magnitude.
  • The Afek initialization is markedly suboptimal for N greater than or equal to 3, so re-optimization of similar setups can recover large fractions of the Heisenberg limit.
  • Post-selection overhead drops sharply, lowering the total photon resources needed to reach a target precision.
  • The inter-channel trade-off identified at N=2 relaxes at higher N, suggesting that multi-channel coincidence analysis becomes more efficient once parameters are tuned.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same gradient-based framework could be applied to optimize probes for multi-parameter estimation or alternative entangled states such as cluster states.
  • Hardware experiments using these parameters would need to verify that unmodeled effects like timing jitter or mode mismatch do not erase the simulated gains.
  • Many existing quantum sensing experiments may improve by numerically re-optimizing their fixed analytical designs rather than building larger N states.
  • Extending the optimization loop to N=6 or higher would test whether the percentage of the Heisenberg limit continues to rise or saturates.

Load-bearing premise

The chosen loss and noise model in the simulation accurately represents real experimental imperfections and the optimized parameters transfer directly to hardware without additional unmodeled effects.

What would settle it

Implementing the Adam-optimized parameters in a physical hybrid coherent-squeezed interferometer and comparing measured phase sensitivity and useful event rate against the Afek baseline at N=3-5.

Figures

Figures reproduced from arXiv: 2604.12323 by Nandan S Bisht, Simanshu Kumar.

Figure 1
Figure 1. Figure 1: Adaptive NOON-state cir￾cuit schematic. The two-mode linear￾optical circuit: state preparation (|α⟩ coher￾ent, |r⟩ squeezed vacuum), input phase ro￾tations R(dcoh) and R(dsq), first beamsplit￾ter BS(θ1, φ1), phase encoding R(φest) on mode 0, and second beamsplitter BS(θ2, φ2). Eight parameters are trainable; φest is scanned to produce coincidence fringes. 2.6 Wigner Function Analysis The Wigner function W(… view at source ↗
Figure 2
Figure 2. Figure 2: Raw CFI comparison: Afek initialisation vs. optimised parameters for all N = 2–5. Grey bars: Afek initialisation. Coloured bars: gradient-optimised. Red dashed line: Heisenberg limit HL= N2 . Percentage labels on optimised bars; colour per N: blue (N = 2), orange (N = 3), green (N = 4), red (N = 5). At N = 2, |1, 1⟩ improves (+153%) while |2, 0⟩ degrades (−59%), reflecting the inter-channel trade-off. At N… view at source ↗
Figure 3
Figure 3. Figure 3: Fringe gallery: Afek initialisation vs. optimised parameters for all N = 2– 5 coincidence patterns. Each row shows one coincidence pattern (N1, N2). Left column: Afek working point. Right column: gradient-optimised parameters. Solid coloured curves: normalised coincidence probability P(N1, N2; φ)/Pmax (left y-axis). Dashed grey curves: classical CFI profile F(φ) (right y-axis, arbitrary scale per panel). T… view at source ↗
Figure 4
Figure 4. Figure 4: Pareto trade-off: fringe quality vs. post-selection rate for N = 2–5. Open circle ◦: Afek initialisation. Star ⋆: optimised. Arrows show the direction of change upon optimisation. x-axis: post-selection rate Pmax. y-axis: fringe quality F norm peak /N2 (Heisenberg limit = 1). (a) N = 2: inter-channel trade-off visible. N = 3: both patterns improve simultaneously. (b) N = 4, 5: large rightward shift (10× hi… view at source ↗
Figure 5
Figure 5. Figure 5: Wigner function gallery of the probe state for all N = 2–5. The probe state is the two-mode state after BS1 but before phase encoding, constituting the non￾classical sensing resource. Each row: one N value. Columns (left to right): mode 0 Afek, mode 0 Optimised, mode 1 Afek, mode 1 Optimised. Colour: blue (W < 0, non-classical), white (W = 0), red (W > 0, classical). White contour: W = 0 boundary. Dotted c… view at source ↗
Figure 6
Figure 6. Figure 6: Wigner negativity N of the probe state: quantitative comparison. (a) Negativity volume for mode 0 (coherent branch), N = 2–5. Grey: Afek; coloured: Opt. (b) Negativity volume for mode 1 (squeezed branch). At N ≥ 3, absolute N increases from ≲ 10−5 (Afek) to 0.003–0.006 (Opt.), confirming genuine enhancement of quantum character. 2. The probe state after BS1 develops interference fringes. For mode 0 after B… view at source ↗
Figure 7
Figure 7. Figure 7: Phase-space portrait of quantum states (N = 2). Heatmap: Wigner function of Afek state. White dashed: Afek constant-W contours; coloured solid: optimised contours; thick solid: W = 0 boundary of the optimised state. (a) Input mode 0: coherent state |α⟩ after R(dcoh). (b) Input mode 1: squeezed vacuum |r⟩; optimisation rotates the squeezing ellipse (r : 0.35 → 0.62). (c,d) Probe modes after BS1: entangled, … view at source ↗
Figure 8
Figure 8. Figure 8: State evolution through the circuit (N = 2): Afek (top) vs. optimised (bottom). Columns: successive circuit stages. Cols 1–2 (before BS1): classical inputs (W ≥ 0); optimised inputs show larger α and r. Central arrow: BS1 (θ1 = π/4). Cols 3–4 (probe state, after BS1): negative Wigner regions (blue) certify non-classicality 25; optimised probe has larger amplitude consistent with improved raw CFI. 6. Analys… view at source ↗
Figure 9
Figure 9. Figure 9: summarises the raw CFI improve￾ment as a function of N. N = 2 N = 3 N = 4 N = 5 Total photon number N 0 500 1000 1500 2000 Raw CFI improvement (%) 1,1 2,0 2,1 3,0 3,1 2,2 3,2 improvement degradation a Raw CFI change N = 2 N = 3 N = 4 N = 5 Total photon number N 0 500 1000 1500 2000 2500 3000 3500 4000 Post-selection rate improvement (%) 1,1 2,0 2,1 3,0 3,1 2,2 3,2 b Post-selection rate change N = 2 N = 3 N… view at source ↗
Figure 10
Figure 10. Figure 10: Optimisation convergence: normalised CFI vs. training step for N = 2–5. Each curve shows the differen￾tiable CFI estimator normalised by its step-0 value (Afek initialisation). All curves con￾verge within 100 Adam steps. Higher N shows larger normalised gain, consistent with the Afek initialisation being increasingly sub￾optimal at higher photon numbers. Training time: ∼ 220 s (N = 2) to ∼ 400 s (N = 5) o… view at source ↗
Figure 11
Figure 11. Figure 11: Parameter drift from the Afek initialisation for N = 2–5. Bars: ∆θi = θ opt i − θ Afek i . The dominant change across all N is an increase in r and log γ, reflecting the optimiser’s strategy of increasing photon flux to improve post-selection rates. Beamsplitter angles show larger drift at N ≥ 3, indicating measurement-basis optimisation is increasingly important at high N. Afek rates. This means variatio… view at source ↗
Figure 12
Figure 12. Figure 12: Marginal photon-number distributions P(n0) for mode 0, N = 2–5. Top: Afek initialisation. Bottom: optimised parameters. Red dashed line: n0 = N (target photon number). After optimisation, distributions broaden toward larger n0, reflecting the increased r and α. This redistribution is the direct mechanism for improved post-selection rates: more probability weight lands in the (N1, N2) coincidence windows. … view at source ↗
Figure 13
Figure 13. Figure 13: Quantum Fisher Information analysis for all N = 2–5. (a) Probe quality FQ/N2 , where FQ = 4 Var(ˆn0) is the exact QFI of the post-BS1 probe state (Eq. 3). Dashed line: Heisenberg limit (= 1); shaded region: classically achievable val￾ues. Percentage labels show the absolute change (pp = percentage points) upon op￾timisation. At N = 3, 4 the optimiser de￾creases probe quality slightly while massively incre… view at source ↗
read the original abstract

Quantum metrology promises phase sensitivity surpassing the shot-noise limit by exploiting entanglement and photon-number correlations. NOON states-maximally path-entangled $N$-photon superpositions $(|N,0\rangle + |0,N\rangle)/\sqrt{2}$ -achieve the Heisenberg limit $1/N$ for single-parameter estimation, as demonstrated experimentally by Afek et al. (2010) using hybrid coherent-plus-squeezed light up to $N=5$. We present an end-to-end differentiable quantum-optical framework-implemented in Strawberry Fields (Killoran et al., 2019) with a TensorFlow backend -that learns optimal circuit parameters by maximising the classical Fisher information (CFI) across all coincidence channels for $N=2,3,4,5$. Starting from proper numerical reproductions of the Afek et al. coincidence fringes, verified by FFT analysis and parity measurements, we apply gradient descent (Adam) to the eight trainable circuit parameters. Raw CFI improvements grow dramatically with photon number: $+153\%$ ($N=2$), $+834\%$ to $+956\%$ ($N=3$), $+829\%$ to $+1598\%$ ($N=4$), and $+1775\%$ ($N=5$), alongside post-selection rate improvements of $+153\%$ to $+3269\%$, and an $8\times$ to $133\times$ improvement in useful measurement events per pulse across $N=2$-$5$. A fundamental inter-channel trade-off is identified at $N=2$ but weakens at higher $N$ where the Afek initialisation is further from optimal. These results provide numerically rigorous benchmarks for adaptive single-parameter quantum sensing and demonstrate that the Afek working point is significantly suboptimal at $N\geq 3$. QFI calculations confirm that the optimised probe reaches $82\%$ of the Heisenberg limit at $N=2$ and improves from $36\%$ to $58\%$ at $N=5$, while useful measurement events per pulse improve by $8\times$ to $133\times$ across all $N$, making quantum-enhanced sensing at $N\geq 3$ experimentally practical.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces an end-to-end differentiable quantum-optical simulation framework in Strawberry Fields with TensorFlow to optimize eight circuit parameters for single-parameter phase estimation using adaptive NOON states at N=2 to 5. Starting from numerical reproductions of Afek et al. (2010) coincidence fringes (verified via FFT and parity), Adam gradient descent maximizes the classical Fisher information (CFI) across coincidence channels, yielding reported CFI gains of +153% (N=2) to +1775% (N=5), post-selection improvements up to +3269%, and 8×–133× more useful events per pulse. QFI analysis shows the optimized states reach 82% of the Heisenberg limit at N=2 and improve from 36% to 58% at N=5, with the conclusion that this renders quantum-enhanced sensing practical for N≥3.

Significance. If the simulation model is faithful, the work provides concrete numerical benchmarks demonstrating that the Afek et al. preparation is substantially suboptimal for N≥3 and that gradient-based optimization of circuit parameters can yield large gains in information extraction and event rates. The differentiable framework itself is a methodological strength, enabling systematic search over parameter space rather than manual tuning, and the reproduction of prior fringes plus separate QFI calculations offer internal consistency checks.

major comments (2)
  1. [Abstract and numerical results] Abstract and numerical results section: The central claim that optimized probes make N≥3 sensing 'experimentally practical' (via 8×–133× more useful events) rests on the Strawberry Fields loss/noise model faithfully capturing dominant experimental imperfections. Only the initial Afek et al. fringes are verified by FFT/parity; no ablation studies, sensitivity analysis to loss parameters, or transfer of optimized parameters to a physical setup are described, leaving the practicality conclusion model-dependent.
  2. [Numerical optimization procedure] Numerical optimization procedure (assumed §3): No details are provided on convergence behavior, number of random initializations, or sensitivity to Adam hyperparameters and learning rate. Since the reported CFI gains are obtained by direct maximization of the same objective, robustness checks are required to confirm the improvements are not artifacts of local optima or initialization bias.
minor comments (1)
  1. [Abstract] Abstract: The phrase 'useful measurement events per pulse improve by 8× to 133× across all N' is repeated verbatim in two consecutive sentences; this redundancy should be removed for clarity.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their thoughtful review and for recognizing the methodological value of the differentiable simulation framework. We address each major comment below. Where the comments identify gaps in the original submission, we have revised the manuscript to provide additional details and clarifications while maintaining that the core numerical results remain valid within the stated model.

read point-by-point responses

  1. Referee: [Abstract and numerical results] Abstract and numerical results section: The central claim that optimized probes make N≥3 sensing 'experimentally practical' (via 8×–133× more useful events) rests on the Strawberry Fields loss/noise model faithfully capturing dominant experimental imperfections. Only the initial Afek et al. fringes are verified by FFT/parity; no ablation studies, sensitivity analysis to loss parameters, or transfer of optimized parameters to a physical setup are described, leaving the practicality conclusion model-dependent.

    Authors: We agree that the practicality statement is framed within the simulation model and that experimental transfer lies beyond the scope of this numerical study. The Strawberry Fields loss model follows standard photonic channel descriptions used in prior works, and our FFT/parity verification of the Afek et al. fringes confirms the base simulation fidelity. To address the concern, we have added a new subsection discussing the model assumptions, including the dominant loss mechanisms, and performed a sensitivity analysis by varying the loss parameters by ±20% around the nominal values; the reported CFI gains remain above +500% for N=5 across this range. We have also tempered the abstract and conclusion language to emphasize that the improvements are demonstrated within the model and provide benchmarks for future experimental implementations rather than claiming direct experimental practicality. revision: partial

  2. Referee: [Numerical optimization procedure] Numerical optimization procedure (assumed §3): No details are provided on convergence behavior, number of random initializations, or sensitivity to Adam hyperparameters and learning rate. Since the reported CFI gains are obtained by direct maximization of the same objective, robustness checks are required to confirm the improvements are not artifacts of local optima or initialization bias.

    Authors: We appreciate this observation and have expanded the methods section with the requested details. The revised manuscript now includes: (i) convergence curves for representative optimizations at each N showing stabilization within 200–400 Adam steps; (ii) results from 20 independent random initializations per N, with the reported CFI values corresponding to the top-performing runs (median gains remain within 10% of the best); and (iii) hyperparameter sweeps over learning rates (0.001–0.05) and β1/β2 values, confirming that the large gains (>800% for N≥3) are reproducible and not sensitive to these choices. These additions demonstrate that the improvements are robust rather than optimization artifacts. revision: yes

standing simulated objections not resolved
  • Direct experimental transfer of the optimized circuit parameters to a physical setup, as the work is a numerical simulation study and hardware implementation is outside its scope.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper reproduces Afek et al. coincidence fringes as baseline, then applies Adam gradient descent to maximize CFI over eight circuit parameters in the Strawberry Fields simulation. Reported CFI gains, post-selection improvements, and QFI-based percentages of the Heisenberg limit are direct numerical outcomes of this optimization and separate QFI evaluation on the resulting probe states. No equations reduce to inputs by construction, no load-bearing self-citations appear, no uniqueness theorems or ansatzes are imported from prior author work, and no known empirical patterns are merely renamed. The derivation chain is self-contained as a demonstration of differentiable optimization within the stated loss/noise model.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on a numerical optimization inside a quantum-optical simulator; the trainable parameters are the only free quantities introduced, while the model itself inherits standard assumptions from quantum optics and the Strawberry Fields library.

free parameters (1)
  • eight trainable circuit parameters
    These parameters are adjusted by gradient descent (Adam) to maximize the classical Fisher information across coincidence channels.
axioms (1)
  • domain assumption The Strawberry Fields simulation accurately captures the linear-optical circuit, photon-number-resolving detectors, and loss model for NOON-state preparation and measurement.
    Invoked when the framework is used to compute CFI and when results are compared to Afek et al. experimental fringes.

pith-pipeline@v0.9.0 · 5709 in / 1526 out tokens · 88642 ms · 2026-05-10T15:07:48.094331+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. OAM-Induced Lattice Rotation Reveals a Fractional Optimum in Fault-Tolerant GKP Quantum Sensing

    quant-ph 2026-05 unverdicted novelty 7.0

    Fractional OAM charge ℓ=1.5 optimizes twisted GKP lattices, cutting error probability by 23.9× versus square lattices at fixed Fisher information.

  2. OAM-Induced Lattice Rotation Reveals a Fractional Optimum in Fault-Tolerant GKP Quantum Sensing

    quant-ph 2026-05 unverdicted novelty 6.0

    Fractional OAM charge ℓ=1.5 induces an optimal 67.5° GKP lattice rotation that reduces error rate 23.9× with <0.2% loss in Fisher information and yields 41% higher metrological capacity.

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages · cited by 1 Pith paper · 2 internal anchors

  1. [1]

    High-NOON states by mixing quantum and classical light.Science328, 879–881 (2010).doi:10.1126/science.1188172

    Afek, I., Ambar, O., & Silberberg, Y. High-NOON states by mixing quantum and classical light.Science328, 879–881 (2010).doi:10.1126/science.1188172

    work page doi:10.1126/science.1188172 2010

  2. [2]

    Caves, C. M. Quantum-mechanical noise in an interferometer.Physi- cal Review D23, 1693–1708 (1981). doi:10.1103/PhysRevD.23.1693 21

    work page doi:10.1103/physrevd.23.1693 1981

  3. [3]

    Killoran, N. et al. Strawberry Fields: A software platform for photonic quan- tum computing.Quantum3, 129 (2019). doi:10.22331/q-2019-03-11-129

    work page doi:10.22331/q-2019-03-11-129 2019

  4. [4]

    Bergholm, V. et al. PennyLane: Automatic differentiation of hy- brid quantum-classical computa- tions.arXiv:1811.04968(2018). doi:10.48550/arXiv.1811.04968

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1811.04968 2018

  5. [5]

    Giovannetti, S

    Giovannetti, V., Lloyd, S., & Maccone, L. Advances in quantum metrology. Nature Photonics5, 222–229 (2011). doi:10.1038/nphoton.2011.35

    work page doi:10.1038/nphoton.2011.35 2011

  6. [6]

    Quantum metrology.Physi- cal Review Letters96, 010401 (2006)

    Giovannetti, V., Lloyd, S., & Mac- cone, L. Quantum metrology.Physi- cal Review Letters96, 010401 (2006). doi:10.1103/PhysRevLett.96.010401

    work page doi:10.1103/physrevlett.96.010401 2006

  7. [7]

    Dowling, J. P. Quantum opti- cal metrology—the lowdown on high-N00N states.Contempo- rary Physics49, 125–143 (2008). doi:10.1080/00107510802091298

    work page doi:10.1080/00107510802091298 2008

  8. [8]

    Lee, H., Kok, P., & Dowling, J. P. A quantum Rosetta Stone for interferometry.Journal of Mod- ern Optics49, 2325–2338 (2002). doi:10.1080/09500340210123173

    work page doi:10.1080/09500340210123173 2002

  9. [9]

    Physical Review Letters85, 2733–2736 (2000).doi:10.1103/PhysRevLett.85.2733

    Boto, A.N.etal.Quantuminterferomet- ric optical lithography: exploiting en- tanglement to beat the diffraction limit. Physical Review Letters85, 2733–2736 (2000).doi:10.1103/PhysRevLett.85.2733

    work page doi:10.1103/physrevlett.85.2733 2000

  10. [10]

    Y., Higgins, B

    Xiang, G. Y., Higgins, B. L., Berry, D. W., Wiseman, H. M., & Pryde, G. J. Entanglement-enhanced measure- ment of a completely unknown opti- cal phase.Nature Photonics5, 43–47 (2011).doi:10.1038/nphoton.2010.268

    work page doi:10.1038/nphoton.2010.268 2011

  11. [11]

    Knill, E., Laflamme, R., & Milburn, G. J. A scheme for efficient quantum computation with linear optics.Nature 409, 46–52 (2001).doi:10.1038/35051009

    work page doi:10.1038/35051009 2001

  12. [12]

    Gerry, C. C. & Hach, E. E. Genera- tion of NOON states via cross-Kerr in- teraction and homodyne measurement. Physical Review A82, 063804 (2010). doi:10.1103/PhysRevA.82.063804

    work page doi:10.1103/physreva.82.063804 2010

  13. [13]

    Mitarai, M

    Mitarai, K., Negoro, M., Kitagawa, M., & Fujii, K. Quantum circuit learning. Physical Review A98, 032309 (2018). doi:10.1103/PhysRevA.98.032309

    work page doi:10.1103/physreva.98.032309 2018

  14. [14]

    V., Schulte, M., Hammerer, K., & Zoller, P

    Kaubruegger, R., Vasilyev, D. V., Schulte, M., Hammerer, K., & Zoller, P. Quantum variational op- timisation of Ramsey interferom- etry and atomic clocks.Physical Review Letters123, 260505 (2019). doi:10.1103/PhysRevLett.123.260505

    work page doi:10.1103/physrevlett.123.260505 2019

  15. [15]

    & Sanders, B

    Hentschel, A. & Sanders, B. C. Machine learning for precise quan- tum measurement.Physical Re- view Letters104, 063603 (2010). doi:10.1103/PhysRevLett.104.063603

    work page doi:10.1103/physrevlett.104.063603 2010

  16. [16]

    Lumino, A. et al. Experimen- tal phase estimation enhanced by machine learning.Physical Re- view Applied10, 044033 (2018). doi:10.1103/PhysRevApplied.10.044033

    work page doi:10.1103/physrevapplied.10.044033 2018

  17. [17]

    Y., Boixo, S., Smelyanskiy, V

    Niu, M. Y., Boixo, S., Smelyanskiy, V. N., & Neven, H. Universal quan- tum control through deep reinforcement learning.npj Quantum Information5, 33 (2019).doi:10.1038/s41534-019-0141-3

    work page doi:10.1038/s41534-019-0141-3 2019

  18. [18]

    Ashida, Z

    Demkowicz-Dobrzański, R., Górecki, W., & Gutˇ a, M. Multi-parameter quantum metrology.Advances in Physics69, 345–435 (2020). doi:10.1080/00018732.2021.1896786 22

    work page doi:10.1080/00018732.2021.1896786 2020

  19. [19]

    K., Ou, Z

    Hong, C. K., Ou, Z. Y., & Mandel, L. Measurement of subpicosecond time in- tervals between two photons by interfer- ence.Physical Review Letters59, 2044 (1987).doi:10.1103/PhysRevLett.59.2044

    work page doi:10.1103/physrevlett.59.2044 2044

  20. [20]

    Paris, M. G. A. Quantum estimation for quantumtechnology.International Jour- nal of Quantum Information7, 125–137 (2009).doi:10.1142/S0219749909004839

    work page doi:10.1142/s0219749909004839 2009

  21. [21]

    Shenoy, R.A

    Helstrom, C. W.Quantum Detection and Estimation Theory. Academic Press, New York (1976).doi:10.1016/S0065- 2539(08)60258-1

    work page doi:10.1016/s0065- 1976

  22. [22]

    Braunstein, S. L. & Caves, C. M. Statistical distance and the geome- try of quantum states.Physical Re- view Letters72, 3439–3443 (1994). doi:10.1103/PhysRevLett.72.3439

    work page doi:10.1103/physrevlett.72.3439 1994

  23. [23]

    Kingma, D. P. & Ba, J. Adam: A method for stochastic opti- mization.arXiv:1412.6980(2014). doi:10.48550/arXiv.1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1412.6980 2014

  24. [24]

    On the quantum correction for thermodynamic equ ilibrium,

    Wigner, E. On the quantum correc- tion for thermodynamic equilibrium. Physical Review40, 749–760 (1932). doi:10.1103/PhysRev.40.749

    work page doi:10.1103/physrev.40.749 1932

  25. [25]

    & Życzkowski, K

    Kenfack, A. & Życzkowski, K. Negativ- ity of the Wigner function as an indi- cator of non-classicality.Journal of Op- tics B: Quantum and Semiclassical Op- tics6, 396–404 (2004).doi:10.1088/1464- 4266/6/10/003

    work page doi:10.1088/1464- 2004

  26. [26]

    Cahill, K. E. & Glauber, R. J. Density operators and quasiprobability distribu- tions.Physical Review177, 1882–1902 (1969).doi:10.1103/PhysRev.177.1882

    work page doi:10.1103/physrev.177.1882 1902

  27. [27]

    M., & Lam, P

    Bradshaw, M., Assad, S. M., & Lam, P. K. A tight Cramér-Rao bound for joint parameter estimation with a pure two-mode squeezed probe.Physics Letters A382, 1213–1219 (2018). doi:10.1016/j.physleta.2018.02.011

    work page doi:10.1016/j.physleta.2018.02.011 2018

  28. [28]

    & Paris, M

    Olivares, S. & Paris, M. G. A. Bayesian estimation in homodyne interferometry. Journal of Physics B: Atomic, Molecular and Optical Physics42, 055506 (2009). doi:10.1088/0953-4075/42/5/055506

    work page doi:10.1088/0953-4075/42/5/055506 2009

  29. [29]

    Bayesian quantum multiphase estimation algorithm.Physical Re- view Applied16, 014035 (2021)

    Gebhart, V., Smerzi, A., & Pezè, L. Bayesian quantum multiphase estimation algorithm.Physical Re- view Applied16, 014035 (2021). doi:10.1103/PhysRevApplied.16.014035

    work page doi:10.1103/physrevapplied.16.014035 2021

  30. [30]

    F., Kok, P., & Dunning- ham, J

    Knott, P., Proctor, T., Hayes, A., Ralph, J. F., Kok, P., & Dunning- ham, J. Local versus global strate- gies in multiparameter estimation.Phys- ical Review A94, 062312 (2016). doi:10.1103/PhysRevA.94.062312

    work page doi:10.1103/physreva.94.062312 2016

  31. [31]

    J., Knott, P

    Proctor, T. J., Knott, P. A., & Dunning- ham, J. A. Multiparameter estimation in networked quantum sensors.Physi- cal Review Letters120, 080501 (2018). doi:10.1103/PhysRevLett.120.080501 23

    work page doi:10.1103/physrevlett.120.080501 2018