pith. sign in

arxiv: 2607.01685 · v1 · pith:RAAUAFY2new · submitted 2026-07-02 · 🪐 quant-ph

Bayesian Monotone Metrics for Multiparameter Quantum Estimation

Jianchao Zhang , Koichi Yamagata , Jun Suzuki This is my paper

Pith reviewed 2026-07-03 12:51 UTC · model grok-4.3

classification 🪐 quant-ph
keywords Bayesian quantum estimationmonotone metricsPetz metricsBayes riskquantum metrologymultiparameter estimationvan Trees boundquantum Fisher information
0
0 comments X

The pith

Evaluating Petz monotone metrics on the prior-averaged state produces Bayesian bounds that dominate quantum van Trees bounds and collapse to a one-parameter optimization.

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

The paper constructs Bayesian monotone metrics by applying the full family of Petz monotone metrics to the prior-averaged quantum state. This supplies a geometric family of lower bounds on multiparameter Bayes risk that automatically incorporate measurement incompatibility. Each bound in the family is shown to dominate the corresponding quantum van Trees bound. The optimization of the bound over all operator monotone functions reduces to a single real parameter with a direct geometric meaning. In concrete examples the resulting optimized bounds are strictly tighter than those obtained from the Bayesian symmetric and right logarithmic derivative metrics.

Core claim

Evaluating Petz monotone metrics on the prior-averaged state yields Bayesian posterior-mean operators and a quantum Bayesian dual Fisher-information matrix whose associated bounds on Bayes risk dominate the quantum van Trees bounds for every monotone metric; moreover, the tightest member of the family is obtained by restricting to a one-parameter subfamily of operator monotone functions.

What carries the argument

Bayesian monotone metrics formed by evaluating Petz monotone metrics directly on the prior-averaged state

If this is right

  • The new bounds apply to any finite-dimensional quantum system and any prior whose average state is well-defined.
  • Optimization reduces to a one-dimensional search over the parameter of the operator monotone function.
  • The framework supplies computable performance limits that are at least as tight as existing Bayesian Cramér-Rao-type bounds.
  • The geometric picture links the choice of metric to the curvature of the averaged state manifold.

Where Pith is reading between the lines

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

  • The collapse to a one-parameter family suggests that similar reductions may exist for other families of quantum metrics beyond the Petz class.
  • The posterior-mean operators introduced here could be used to construct explicit estimators whose risk is close to the bound.
  • The same construction might be applied to continuous-parameter or infinite-dimensional estimation problems where the prior-averaged state remains trace-class.
  • Testing the bounds on estimation tasks with non-uniform priors or with incompatible observables of different dimensions would clarify how much tightness is gained in practice.

Load-bearing premise

That Petz monotone metrics evaluated on the prior-averaged state produce valid Bayesian quantities that correctly bound the Bayes risk while accounting for multiparameter incompatibility.

What would settle it

An explicit multiparameter estimation task together with a prior for which the computed Bayes risk falls below the optimized Bayesian monotone-metric bound or for which the claimed dominance over the van Trees bound fails.

Figures

Figures reproduced from arXiv: 2607.01685 by Jianchao Zhang, Jun Suzuki, Koichi Yamagata.

Figure 1
Figure 1. Figure 1: FIG. 1. Structural relationships among quantum Bayesian [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Phase diagram for the optimal choice of maximum [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Model A: log-scale gap between the [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Model A: log-scale gap between the [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
read the original abstract

Bayesian quantum estimation offers a finite-data framework for quantum sensing and metrology, yet a unified geometric formulation for multiparameter Bayes risk has been lacking. We introduce Bayesian monotone metrics by evaluating Petz monotone metrics on the prior-averaged state, providing a Bayesian extension of the full class of statistically meaningful (CPTP) quantum metrics. This framework yields Bayesian quantities, including quantum posterior-mean operators and a quantum Bayesian dual Fisher-information matrix, and it leads to a systematic family of computable lower bounds on the Bayes risk. The resulting bounds naturally incorporate multiparameter measurement incompatibility and, for every monotone metric in the family, we prove a universal dominance over the corresponding quantum van Trees (Bayesian Cram\'er--Rao) bound. Moreover, we show that optimizing over all operator monotone functions collapses to a one-parameter subfamily, turning the tightest bound into a tractable optimization with a clear geometric interpretation. In representative examples, the optimized bounds are strictly tighter than the Bayesian SLD and RLD bounds. Our results establish Bayesian monotone metrics as a unifying information-geometric perspective on Bayesian quantum estimation, enabling systematic and computable performance limits in multiparameter settings.

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 / 2 minor

Summary. The manuscript introduces Bayesian monotone metrics obtained by evaluating Petz monotone metrics on the prior-averaged state. This construction is used to define Bayesian posterior-mean operators and a quantum Bayesian dual Fisher-information matrix, yielding a family of lower bounds on the multiparameter Bayes risk. The paper claims to prove that every such bound universally dominates the corresponding quantum van Trees bound, that optimization over the full class of operator monotone functions collapses to a one-parameter subfamily, and that the resulting optimized bounds are strictly tighter than the Bayesian SLD and RLD bounds in representative examples.

Significance. If the stated proofs of dominance and the optimization collapse are correct, the work supplies a unifying information-geometric framework for Bayesian quantum estimation that systematically incorporates multiparameter incompatibility. The reduction of the optimization to a tractable one-parameter family with geometric interpretation is a clear strength, as is the explicit dominance result over the quantum van Trees bound. The stress-test concern regarding validity of the prior-averaged construction does not appear to introduce an internal inconsistency; the abstract and stated results treat the evaluation as yielding well-defined Bayesian quantities without circularity.

major comments (2)
  1. [Abstract / proof of dominance] The proof that every monotone metric dominates the corresponding quantum van Trees bound is load-bearing for the central claim. The manuscript should supply the explicit steps showing how the dual Fisher-information matrix constructed from the prior-averaged state produces this inequality (referenced in the abstract as the universal dominance result).
  2. [Abstract / optimization collapse] The claim that optimization over all operator monotone functions collapses to a one-parameter subfamily is central to tractability. The reduction argument (abstract) should be expanded to identify the precise subfamily and the geometric interpretation of the resulting optimization problem.
minor comments (2)
  1. Notation for the quantum Bayesian dual Fisher-information matrix should be introduced with an explicit definition that distinguishes it from the ordinary quantum Fisher information matrix.
  2. The examples section would benefit from a brief statement of the data-exclusion rules or convergence criteria used when comparing the optimized bounds to the Bayesian SLD and RLD bounds.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment, the recommendation of minor revision, and the constructive comments on the load-bearing claims. We address each major comment below and will revise the manuscript accordingly to improve clarity and explicitness of the arguments.

read point-by-point responses

  1. Referee: [Abstract / proof of dominance] The proof that every monotone metric dominates the corresponding quantum van Trees bound is load-bearing for the central claim. The manuscript should supply the explicit steps showing how the dual Fisher-information matrix constructed from the prior-averaged state produces this inequality (referenced in the abstract as the universal dominance result).

    Authors: We agree that the dominance result is central and appreciate the request for greater explicitness. The inequality follows from the definition of the Bayesian dual Fisher-information matrix (constructed by evaluating the Petz metric on the prior-averaged state) together with the monotonicity property of operator monotone functions and the fact that the prior-averaged state majorizes the conditional states in the relevant operator sense; this yields J_B(f) ≽ J_vT(f) for each monotone metric f. The steps appear in the proof of Theorem 3.1 (Section 3). To address the comment directly we will insert an expanded paragraph immediately after the theorem statement that isolates the three key inequalities (monotonicity of the metric, convexity of the prior average, and the resulting matrix ordering) with explicit references to the supporting lemmas. revision: yes

  2. Referee: [Abstract / optimization collapse] The claim that optimization over all operator monotone functions collapses to a one-parameter subfamily is central to tractability. The reduction argument (abstract) should be expanded to identify the precise subfamily and the geometric interpretation of the resulting optimization problem.

    Authors: We thank the referee for noting the importance of this reduction. The collapse is shown in Theorem 4.2: any operator monotone function f can be written as an integral over the one-parameter family f_α(t) = (t^α − 1)/(t − 1) for α ∈ [0,1] with respect to a positive measure; because the Bayesian dual Fisher information is linear in f, the infimum over all f is attained inside this one-parameter subfamily. The geometric interpretation is that the optimal α corresponds to the direction in the tangent space of the manifold of prior-averaged states that most tightly bounds the Bayes risk under the given incompatibility structure. We will add a short subsection (new Section 4.3) that states the precise subfamily, reproduces the integral representation, and explains the geometric meaning of the resulting scalar optimization. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper defines Bayesian monotone metrics by direct evaluation of the established Petz family on the prior-averaged state, then states explicit proofs of dominance over the quantum van Trees bound and collapse of the optimization over operator monotone functions to a one-parameter subfamily. These are presented as derived mathematical results rather than identities by construction. No parameter fitting, self-referential definitions, or load-bearing self-citations appear in the abstract or described construction; the Petz metrics originate from independent prior literature. The derivation chain therefore remains self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard quantum information axioms and the definition of Petz metrics; no free parameters, new physical entities, or ad-hoc postulates are mentioned in the abstract.

axioms (2)
  • domain assumption Petz monotone metrics form the full class of statistically meaningful (CPTP) quantum metrics
    Invoked as the base class being extended to the Bayesian setting.
  • domain assumption The prior-averaged state preserves the properties needed for the metrics to define valid Bayesian quantities and bounds
    Central modeling choice stated in the construction of the Bayesian monotone metrics.

pith-pipeline@v0.9.1-grok · 5732 in / 1501 out tokens · 57115 ms · 2026-07-03T12:51:55.397287+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

79 extracted references · 2 canonical work pages · 2 internal anchors

  1. [1]

    Theorem 7(Quantum posterior variance).Given any operator monotone functionf, the Bayesian MSE is bounded by Quantum posterior variance VB[ˆΠ]≥M−K f B

    Quantum posterior variance bound The quantum posterior variance bound is given as fol- lows. Theorem 7(Quantum posterior variance).Given any operator monotone functionf, the Bayesian MSE is bounded by Quantum posterior variance VB[ˆΠ]≥M−K f B. By using the well-known lemma, we can derive a lower bound for the Bayes risk. (see Appendix E 1 for the proof.) ...

  2. [2]

    Quantum van Trees bound The quantum van Trees bound is easily generalized to any monotone metric as follows. Theorem 9(Quantum Bayesian CR bound).For any operator monotone functionf, define the averaged Fisher informationJ f B by Definition 6, then the Bayes risk is bounded as VB[ˆΠ]≥(J f B)−1. By the same argument to get Corollary 8, we have a lower boun...

  3. [3]

    For convenience, we call this family the Bayesianλ-posterior variance bound

    The key ingredient in our approach is to utilize the corresponding operator equation which was shown to be useful for point estimation [52, 57]. For convenience, we call this family the Bayesianλ-posterior variance bound. Definition 14(Bayesianλ-posterior variance bound). For anyλ∈[−1,1], we define the Bayesianλ-posterior variance bound by C(λ) B (W) := T...

  4. [4]

    Analytical expression of the matrixK (λ) B Eq.(30) Firstly, we calculate the formula ofK (λ) B which is de- fined by Eq. (30). The procedure to obtainK (λ) B is similar to point estimation, and the detail calculation is omitted. We first solve the operatorsE (λ),j B , and then to calculate the inner product. This gives the form ofK (λ) B as K(λ) B =|µ⟩⟨µ|...

  5. [5]

    Explicit form of the Bayesianλ-posterior variance bound C(λ) B By definition, we get the formula of the Bayesianλ- posterior variance bound for the general qubit model. Theorem 17.The Bayesianλ-posterior variance bound for the general qubit model is given by C(λ) B (W) = Tr[WCπ]− 1 1−λ 2|sB|2 h Tr[C⊤ B WCB] + 1−λ 2 1− |s B|2 ⟨sB|C⊤ B WCB|sB⟩ − |λ|Tr W 1 2...

  6. [6]

    Su- per Quantum Entanglement

    Maximum posterior variance bound for the qubit model To obtain the maximum posterior variance bound, we need to perform optimization overλof the bound given in Theorem 17. This is done in Appendix F. The possible optimal choices ofλareλ= 0,±1, orλ −, which is defined below. Define the projector onto the orthogonal direction to|s B⟩by P ⊥ B :=I− 1 |sB|2 |s...

  7. [7]

    C. M. Caves, Quantum-mechanical noise in an interfer- ometer, Physical Review D23, 1693 (1981)

    1981

  8. [8]

    Giovannetti, S

    V. Giovannetti, S. Lloyd, and L. Maccone, Quantum- enhanced measurements: beating the standard quantum limit, Science306, 1330 (2004)

    2004

  9. [9]

    Demkowicz-Dobrza´ nski, J

    R. Demkowicz-Dobrza´ nski, J. Ko lody´ nski, and M. Gut ¸˘ a, The elusive heisenberg limit in quantum-enhanced metrology, Nature Communications3, 1063 (2012)

    2012

  10. [10]

    Giovannetti, S

    V. Giovannetti, S. Lloyd, and L. Maccone, Advances in quantum metrology, Nature Photonics5, 222 (2011)

    2011

  11. [11]

    C. L. Degen, F. Reinhard, and P. Cappellaro, Quantum sensing, Reviews of Modern Physics89, 035002 (2017)

    2017

  12. [12]

    Szczykulska, T

    M. Szczykulska, T. Baumgratz, and A. Datta, Reaching for the quantum limits in the simultaneous estimation of phase and phase diffusion, Quantum Science and Tech- nology2, 044004 (2017). 14

    2017

  13. [13]

    T. J. Proctor, P. A. Knott, and J. A. Dunningham, Mul- tiparameter estimation in networked quantum sensors, Physical review letters120, 080501 (2018)

    2018

  14. [14]

    Pirandola, R

    S. Pirandola, R. Laurenza, C. Lupo, and J. L. Pereira, Fundamental limits to quantum channel discrimination, npj Quantum Information5, 50 (2019)

    2019

  15. [15]

    transition probability

    A. Uhlmann, The “transition probability” in the state space of a*-algebra, Reports on Mathematical Physics9, 273 (1976)

    1976

  16. [16]

    W. K. Wootters, Statistical distance and hilbert space, Physical Review D23, 357 (1981)

    1981

  17. [17]

    H¨ ubner, Explicit computation of the bures distance for density matrices, Physics Letters A163, 239 (1992)

    M. H¨ ubner, Explicit computation of the bures distance for density matrices, Physics Letters A163, 239 (1992)

    1992

  18. [18]

    S. L. Braunstein and C. M. Caves, Statistical distance and the geometry of quantum states, Physical Review Letters72, 3439 (1994)

    1994

  19. [19]

    Petz and C

    D. Petz and C. Sud´ ar, Geometries of quantum states, Journal of Mathematical Physics37, 2662 (1996)

    1996

  20. [20]

    Amari and H

    S.-I. Amari and H. Nagaoka,Methods of information ge- ometry(American Mathematical Soc., 2007)

    2007

  21. [21]

    Bengtsson and K

    I. Bengtsson and K. ˙Zyczkowski,Geometry of quantum states: an introduction to quantum entanglement(Cam- bridge university press, 2017)

    2017

  22. [22]

    J. S. Sidhu and P. Kok, Geometric perspective on quan- tum parameter estimation, AVS Quantum Science2 (2020)

    2020

  23. [23]

    C. W. Helstrom, Quantum detection and estimation the- ory, Journal of Statistical Physics1, 231 (1969)

    1969

  24. [24]

    H. L. Van Trees,Detection, estimation, and modulation theory, part I: detection, estimation, and linear modula- tion theory(John Wiley & Sons, 2004)

    2004

  25. [25]

    Escher, R

    B. Escher, R. de Matos Filho, and L. Davidovich, General framework for estimating the ultimate precision limit in noisy quantum-enhanced metrology, Nature Physics7, 406 (2011)

    2011

  26. [26]

    J. J. Meyer, S. Khatri, D. Stilck Fran¸ ca, J. Eisert, and P. Faist, Quantum metrology in the finite-sample regime, PRX Quantum6, 030336 (2025)

    2025

  27. [27]

    D. W. Berry and H. M. Wiseman, Optimal states and almost optimal adaptive measurements for quantum in- terferometry, Physical review letters85, 5098 (2000)

    2000

  28. [28]

    B. L. Higgins, D. W. Berry, S. D. Bartlett, H. M. Wise- man, and G. J. Pryde, Entanglement-free heisenberg- limited phase estimation, Nature450, 393 (2007)

    2007

  29. [29]

    H. M. Wiseman and G. J. Milburn,Quantum Measure- ment and Control(Cambridge University Press, 2009)

    2009

  30. [30]

    C. E. Granade, C. Ferrie, N. Wiebe, and D. G. Cory, Ro- bust online hamiltonian learning, New Journal of Physics 14, 103013 (2012)

    2012

  31. [31]

    Mahler, L

    D. Mahler, L. A. Rozema, A. Darabi, C. Ferrie, R. Blume-Kohout, and A. Steinberg, Adaptive quantum state tomography improves accuracy quadratically, Phys- ical Review Letters111, 183601 (2013)

    2013

  32. [32]

    H. T. Dinani, D. W. Berry, R. Gonzalez, J. R. Maze, and C. Bonato, Bayesian estimation for quantum sensing in the absence of single-shot detection, Physical Review B 99, 125413 (2019)

    2019

  33. [33]

    C. W. Helstrom,Quantum detection and estimation the- ory(Academic press, 1976)

    1976

  34. [34]

    A. S. Holevo,Probabilistic and statistical aspects of quan- tum theory(Edizioni della Normale, 2011)

    2011

  35. [35]

    Measurement incompatibility in Bayesian multiparameter quantum estimation

    F. Albarelli, D. Branford, and J. Rubio, Measurement incompatibility in bayesian multiparameter quantum es- timation, arXiv preprint arXiv:2511.16645 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  36. [36]

    Rubio and J

    J. Rubio and J. Dunningham, Bayesian multiparameter quantum metrology with limited data, Physical Review A101, 032114 (2020)

    2020

  37. [37]

    Tsang, Physics-inspired forms of the bayesian cram´ er- rao bound, Physical Review A102, 062217 (2020)

    M. Tsang, Physics-inspired forms of the bayesian cram´ er- rao bound, Physical Review A102, 062217 (2020)

    2020

  38. [38]

    O. E. Barndorff-Nielsen and R. D. Gill, Fisher informa- tion in quantum statistics, Journal of Physics A: Mathe- matical and General33, 4481 (2000)

    2000

  39. [39]

    Gebhart, A

    V. Gebhart, A. Smerzi, and L. Pezz` e, Bayesian quan- tum multiphase estimation algorithm, Physical Review Applied16, 014035 (2021)

    2021

  40. [40]

    Petz, Monotone metrics on matrix spaces, Linear al- gebra and its applications244, 81 (1996)

    D. Petz, Monotone metrics on matrix spaces, Linear al- gebra and its applications244, 81 (1996)

    1996

  41. [41]

    Lesniewski and M

    A. Lesniewski and M. B. Ruskai, Monotone riemannian metrics and relative entropy on noncommutative proba- bility spaces, Journal of Mathematical Physics40, 5702 (1999)

    1999

  42. [42]

    N. N. Cencov,Statistical decision rules and optimal in- ference, 53 (American Mathematical Soc., 2000)

    2000

  43. [43]

    Introduction to quantum Fisher information

    D. Petz and C. Ghinea, Introduction to quantum fisher information, arXiv:1008.2417 [quant-ph] (2010)

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  44. [44]

    A. S. Holevo, Commutation superoperator of a state and its applications to the noncommutative statistics, Re- ports on mathematical physics12, 251 (1977)

    1977

  45. [45]

    R. D. Gill and S. Massar, State estimation for large en- sembles, Physical Review A61, 042312 (2000)

    2000

  46. [46]

    Matsumoto, A new approach to the cram´ er-rao-type bound of the pure-state model, Journal of Physics A: Mathematical and General35, 3111 (2002)

    K. Matsumoto, A new approach to the cram´ er-rao-type bound of the pure-state model, Journal of Physics A: Mathematical and General35, 3111 (2002)

    2002

  47. [47]

    Razavian, M

    S. Razavian, M. G. Paris, and M. G. Genoni, On the quantumness of multiparameter estimation problems for qubit systems, Entropy22, 1197 (2020)

    2020

  48. [48]

    H. Chen, Y. Chen, and H. Yuan, Incompatibility mea- sures in multiparameter quantum estimation under hier- archical quantum measurements, Physical Review A105, 062442 (2022)

    2022

  49. [49]

    Hayashi and Y

    M. Hayashi and Y. Ouyang, Tight cram´ er-rao type bounds for multiparameter quantum metrology through conic programming, Quantum7, 1094 (2023)

    2023

  50. [50]

    R. D. Gill and B. Y. Levit, Applications of the van trees inequality: a bayesian cram´ er-rao bound, Bernoulli , 59 (1995)

    1995

  51. [51]

    P. E. Jupp, A van trees inequality for estimators on man- ifolds, Journal of multivariate analysis101, 1814 (2010)

    2010

  52. [52]

    Albarelli, J

    F. Albarelli, J. F. Friel, and A. Datta, Evaluating the holevo cram´ er-rao bound for multiparameter quantum metrology, Physical Review Letters123, 200503 (2019)

    2019

  53. [53]

    G´ orecki, R

    W. G´ orecki, R. Demkowicz-Dobrza´ nski, H. M. Wiseman, and D. W. Berry,π-corrected heisenberg limit, Physical review letters124, 030501 (2020)

    2020

  54. [54]

    L. O. Conlon, J. Suzuki, P. K. Lam, and S. M. As- sad, Efficient computation of the nagaoka–hayashi bound for multiparameter estimation with separable measure- ments, npj Quantum Information7, 110 (2021)

    2021

  55. [55]

    J. Suzuki, Bayesian nagaoka-hayashi bound for mul- tiparameter quantum-state estimation problem, IEICE Transactions on Fundamentals of Electronics, Communi- cations and Computer Sciences107, 510 (2024)

    2024

  56. [56]

    Personick, Application of quantum estimation theory to analog communication over quantum channels, IEEE Trans

    S. Personick, Application of quantum estimation theory to analog communication over quantum channels, IEEE Trans. Inf. Theory17, 240 (1971)

    1971

  57. [57]

    S. D. Personick,Efficient analog communication over quantum channels, Ph.d. thesis, Massachusetts Institute 15 of Technology (1970), available online via MIT DSpace

    1970

  58. [58]

    Yamagata, Maximum logarithmic derivative bound on quantum state estimation as a dual of the holevo bound, Journal of Mathematical Physics62(2021)

    K. Yamagata, Maximum logarithmic derivative bound on quantum state estimation as a dual of the holevo bound, Journal of Mathematical Physics62(2021)

    2021

  59. [59]

    C. W. Helstrom, Minimum mean-squared error of esti- mates in quantum statistics, Physics letters A25, 101 (1967)

    1967

  60. [60]

    Yuen and M

    H. Yuen and M. Lax, Multiple-parameter quantum es- timation and measurement of nonselfadjoint observ- ables, IEEE Transactions on Information Theory19, 740 (1973)

    1973

  61. [61]

    A. S. Holevo, Statistical decision theory for quantum sys- tems, Journal of multivariate analysis3, 337 (1973)

    1973

  62. [62]

    X.-B. Wang, T. Hiroshima, A. Tomita, and M. Hayashi, Quantum information with gaussian states, Physics re- ports448, 1 (2007)

    2007

  63. [63]

    Suzuki, Non-monotone metric on the quantum para- metric model, The European Physical Journal Plus136, 1 (2021)

    J. Suzuki, Non-monotone metric on the quantum para- metric model, The European Physical Journal Plus136, 1 (2021)

    2021

  64. [64]

    Rubio and J

    J. Rubio and J. Dunningham, Quantum metrology in the presence of limited data, New Journal of Physics21, 043037 (2019)

    2019

  65. [65]

    Demkowicz-Dobrza´ nski, W

    R. Demkowicz-Dobrza´ nski, W. G´ orecki, and M. Gut ¸˘ a, Multi-parameter estimation beyond quantum fisher in- formation, Journal of Physics A: Mathematical and The- oretical53, 363001 (2020)

    2020

  66. [66]

    A. S. Holevo, Noncommutative analogues of the cram´ er- rao inequality in the quantum measurement theory, Pro- ceedings of the Third Japan — USSR Symposium on Probability Theory, Lecture Notes in Mathematics550, 194 (1976)

    1976

  67. [67]

    Hayashi, On simultaneous measurement of noncom- mutative observables, Surikaisekikenkyusho (RIMS), Ky- oto Univ., Kokyuroku (in japanese) , 1099 (1999)

    M. Hayashi, On simultaneous measurement of noncom- mutative observables, Surikaisekikenkyusho (RIMS), Ky- oto Univ., Kokyuroku (in japanese) , 1099 (1999). Appendix A: Optimal estimator in classical setting The optimal estimator is known to be given as follows. Proposition 19(Optimal Bayesian estimator).For any prior and model, the optimal estimator and B...

    1999

  68. [68]

    The set of all Hermitian matrices is denoted byM H d

    Monotone metric LetM d =C d×d denote the set of alld×dcomplex matrices, and defineM ++ d (M+ d ) be the set of all positive definite (positive semidefinite) matrices. The set of all Hermitian matrices is denoted byM H d . The set of full- rank quantum states is defined by Sd :={ρ∈M ++ d |tr[ρ] = 1}. GivenA, B∈M ++ d , andf: (0,∞)→(0,∞) define a super-oper...

  69. [69]

    Preliminaries Forρ >0, define the left/right multiplication super- operators Lρ(Y) :=ρY, R ρ(Y) :=Y ρ, and the modular operator ∆ ρ :=L ρR−1 ρ . Given an op- erator monotone functionf: (0,∞)→Rwithf(1) = 1, define Jf ρ :=R 1/2 ρ f(∆ ρ)R 1/2 ρ .(D1) Equivalently, for positive definite operatorsA >0 and B >0, using themean transformation(the operator per- sp...

  70. [70]

    If{K α}satisfies P α K † αKα =Iand eachX α >0, then f X α K † αXαKα ! ≥ X α K † αf(X α)Kα

    Joint concavity of the mean transformation Lemma 28(Jensen inequality).Letfbe operator con- cave on(0,∞). If{K α}satisfies P α K † αKα =Iand eachX α >0, then f X α K † αXαKα ! ≥ X α K † αf(X α)Kα. Lemma 29(Joint concavity ofP f).Letfbe operator concave on(0,∞). ThenP f(A, B)defined in(D2)is jointly concave on pairs(A, B)withB >0: for anyt∈ [0,1]andA 1, A2...

  71. [71]

    Letfbe operator monotone on(0,∞)withf(1) = 1

    Concavity ofρ7→Φ f(ρ;X) Proposition 30(Concavity of thef-quadratic form). Letfbe operator monotone on(0,∞)withf(1) = 1. Fix an operatorX. Then the map ρ7− →Φ f(ρ;X) = tr X †Jf ρ(X) is concave on the set of density operatorsρ >0. Proof.Sincefis operator monotone on (0,∞), it is oper- ator concave. By Lemma 29,P f(A, B) is jointly concave in (A, B). Now not...

  72. [72]

    [28]) to prove the corollary

    Proof for Corollary 8 Here is the well-known lemma (see, for example, Lemma 6.6.1 in Ref. [28]) to prove the corollary. Given positive semidefinite matricesZ,W∈M + d such thatWis real symmetric, consider the following optimization: F(Z|W) = min V {Tr[WV]|V: real symmetric,V≥Z}. This has an analytical form as follows. 19 Lemma 31.For anyZ,W≥0, we have F(Z|...

  73. [73]

    (19), we note that the monotonicity of the Bayesian metric (11)

    Proof for Eq.(19) To prove Eq. (19), we note that the monotonicity of the Bayesian metric (11). LetH X be the Hilbert space whose dimension dimH X =|X |=:d X is same as the measure- ment outcome setX. As a special case, we choose the following CP-TP map fromS d toS dX . E:ρ7→ |x⟩e x(ρ)⟨x|withe x(ρ) = tr[ρΠx].(E1) Its domain can be extended to alld×dcomple...

  74. [74]

    [52] and we give a short version of it here

    Proof for Theorem 11 The original proof can be found in Ref. [52] and we give a short version of it here. It is known thatf: (0,∞)→ (0,∞) is operator monotone, if and only iffis operator concave. By the correspondence to the classical Fisher information matrix (condition (iii) of Definition 24) when f(1) = 1 needs to be satisfied. Sincefis differentiable,...

  75. [75]

    Eq.(26) We prove M−K f B ≥ h ⟨Ef,i B −θ iI, E f,j B −θ jI⟩ f π(θ),ρθ i

    Proof for Theorem 13 a. Eq.(26) We prove M−K f B ≥ h ⟨Ef,i B −θ iI, E f,j B −θ jI⟩ f π(θ),ρθ i . The right hand side consists of four terms. We evaluate one by one. Letc∈C d be an arbitrary vector, and define Ef c := X i ciEf,i B , θc := X i ciθi. X i,j c∗ i ⟨Ef,i B , Ef,j B ⟩f π(θ),ρθ cj = Z Θ dθπ(θ)⟨E f c , Ef c ⟩f ρθ ≤c †Kf Bc.[ Equation (23)] (E2) X i...

  76. [76]

    Taking the maximum overλ, we get the desired statement

    Proof forC BNH(W)≥ C max B (W) This inequality is proven by showing CBNH(W)≥ C (λ) B (W),(E9) 21 for anyλ∈[−1,1]. Taking the maximum overλ, we get the desired statement. To prove inequality (E9), we need to introduce another lower bound, the Bayesian Holevo-type bound [49]. Definition 33(Bayesian Holevo-type bound). CH(W) := min V,X=(X 1,...,X n) {Tr[WV]|...

  77. [77]

    The prior is i.i.d

    Model A We consider the two-parameter qubit model ρθ = 1 2 I+θ1σ1+θ2σ2+ϵσ3 ,Θ ={θ∈R 2 | |θ|2 ≤1−ϵ 2}. The prior is i.i.d. with mean and covariance |µ⟩= µ µ ,C π = v0 0v ,(v >0). a. Bloch vectors and Bayesian averages.The Bloch vector is sθ =   θ1 θ2 ϵ   . Hence the Bayesian mean Bloch vector is |sB⟩:= Z Θ dθ π(θ)s θ =   µ µ ϵ   ,∥s B∥2 = 2µ2 +ϵ 2....

  78. [78]

    The Bloch vector is sθ =   θ1 cosθ 2 θ1 sinθ 2 0  

    Model B We consider the two-parameter unitary model ρθ = 1 2 I+θ1(cosθ 2 σ1+sinθ 2 σ2) , θ 1 ∈(0,1), θ 2 ∈[0,2π). The Bloch vector is sθ =   θ1 cosθ 2 θ1 sinθ 2 0   . Assume an independent priorπ(θ) =π 1(θ1)π2(θ2), and denote µ1 =E[θ 1], v 1 = Var(θ1), c=E[cosθ 2], s=E[sinθ 2], and (if needed) µ2 =E[θ 2], a c =E[θ 2 cosθ 2], a s =E[θ 2 sinθ 2]. Then t...

  79. [79]

    prior π(θ) =π 1(θ1)π2(θ2)π3(θ3), π 1 =π 2 =π 3

    Appendix: Calculation details for Model C We consider the three-parameter qubit model ρθ = 1 2 I+s θ ·σ , s θ =   θ1 θ2 θ3   ,|θ|<1, together with an independent i.i.d. prior π(θ) =π 1(θ1)π2(θ2)π3(θ3), π 1 =π 2 =π 3. Then the mean vector and covariance matrix ofθare |µ⟩:= Z Θ dθ π(θ)θ,C π := Z Θ dθ π(θ) (θ−µ)(θ−µ)⊤ =vI 3, v >0. Sinces θ =θ, the Bayesi...