Pith. sign in

REVIEW 3 major objections 3 minor 109 references

Unifying quantum measurement constructions via a relative-entropy minimum change principle

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

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

Pith's one-line read A relative-entropy minimum change principle determines the optimal quantum measurement in closed form.

desk verdict A clean variational principle that unifies PGMs and thermal measurements, with a caveat that the full-rank assumption is load-bearing. read the letter →

arxiv 2608.04055 v1 pith:USCEUV5C submitted 2026-08-04 quant-ph cond-mat.stat-mechcs.ITcs.LGmath.IT

Nana Liu , Mark M. Wilde This is my paper · ORCID
classification quant-phcond-mat.stat-mechcs.ITcs.LGmath.IT MSC 81P4581P5094A17
keywords quantumrelativeentropyminimumchangeprinciplePOVMprettygoodmeasurementFermi-Diracthermalsoftminhypothesistestingsemidefiniteoptimization
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 aims to show that a single variational principle—minimize the quantum relative entropy between a forward classical-to-quantum preparation process and a reverse quantum-to-classical measurement process—determines the optimal measurement in closed form. The solution is a POVM built from a reference state $\tau$ and the unnormalized prior-and-likelihood operators $\sigma_x$, with the reference state as the only free ingredient. By choosing $\tau$ to be the average output state, the principle reproduces the pretty good measurement; by choosing $\tau$ maximally mixed it yields a new softmin thermal measurement, whose binary case is the Fermi–Dirac thermal measurement. If the theorem is right, seemingly distinct measurement recipes are special cases of one information-theoretic optimization, and the softmin thermal measurement also emerges independently as the optimizer of entropy-regularized semidefinite programs.

What carries the argument

The load-bearing object is the dual optimization over a single unconstrained Hermitian variable $A$. After substituting $N_x = \tau^{1/2} M_x \tau^{1/2}$, the constraint that the POVM sums to the identity becomes $\sum_x N_x = \tau$, and the relative-entropy minimization splits into a sum of generalized relative entropies; the Lagrange multiplier $A$ enforces that sum constraint. The stationarity condition produces the operator equation $\sum_x e^{\ln\sigma_x - A} = \tau$, and the optimal POVM recomposes from the same exponentials. The strict convexity of $A \mapsto \mathrm{Tr}[A\tau] + \sum_x \mathrm{Tr}[e^{\ln\sigma_x - A}]$ guarantees uniqueness and underlies the convergence of the gradient iteration.

What would settle it

Take a two-dimensional example with pure $\rho_0$ and mixed $\rho_1$, so $\sigma_0$ has a zero eigenvalue, and compute both sides of Theorem 1 by numerically optimizing the relative entropy over all POVMs; if the optimized value falls strictly below the right-hand side of (23) evaluated with the formula, the positive-definiteness assumption is genuinely load-bearing, while agreement would indicate the formula extends to the boundary.

Watch

Extended reading notes

Core claim

Theorem 1 states that for positive definite $\tau$ and $\sigma_x$ the minimal change is $\min_{M_X} D(Q_{\mathrm{rev}}(M_X)\|Q_{\mathrm{fwd}}) = 1 - \inf_{A\in\mathrm{Herm}}(\mathrm{Tr}[A\tau] + \sum_x \mathrm{Tr}[e^{\ln\sigma_x - A}])$, with the infimum attained at the unique Hermitian $A^*$ solving $\sum_x e^{\ln\sigma_x - A^*} = \tau$, and the unique optimal POVM given by $M^*_x = \tau^{-1/2} e^{\ln\sigma_x - A^*} \tau^{-1/2}$. Here $Q_{\mathrm{fwd}} = \sum_x \sigma_x \otimes |x\rangle\langle x|$ and $Q_{\mathrm{rev}}(M_X) = \sum_x \tau^{1/2} M_x \tau^{1/2} \otimes |x\rangle\langle x|$, so the relative entropy decouples into a direct sum. The authors show the dual objective is strictly convex, hence the optimizer is unique, and they give a gradient-descent scheme for finding $A^*$. From this single formula they recover the pretty good measurement ($\tau=\sigma$), the Fermi–Dirac thermal measurement (binary alphabet, $\tau=I/d$), and define the softmin thermal measurement ($\tau=I/d$, general alphabet). The paper further proves that softmin thermal measurements are exactly the optimizers of a broad class of entropy-regularized semidefinite programs, that the minimum-change value is additive over product forward processes with product reference states, and that in binary hypothesis testing the Fermi–Dirac thermal measurement's asymptotic error exponent falls strictly below the quantum Chernoff bound for noncommuting states.

Load-bearing premise

The whole characterization assumes the reference state $\tau$ and every substate $\sigma_x = p(x)\rho_x$ are strictly positive definite; if any of them has a zero eigenvalue, the logarithms, inverse square roots, and matrix exponentials in the formula are not defined, and the paper gives no limiting argument for these boundary cases.

Editorial extensions

If this is right

  • Choosing $\tau$ to be the average output state $\sigma$ reproduces the pretty good measurement, with the minimum-change value equal to zero.
  • Choosing $\tau$ maximally mixed defines the softmin thermal measurement; for binary alphabets it reduces to the Fermi–Dirac thermal measurement, so the principle supplies a unified derivation of both.
  • The reference state $\tau$ parametrizes a continuous family of optimal measurements; a thermal interpolation $\tau = e^{-\beta H}/\mathrm{Tr}[e^{-\beta H}]$ with $H=-\ln\sigma$ interpolates between the softmin thermal measurement at $\beta=0$ and the pretty good measurement at $\beta=1$.
  • The additivity theorem implies that for product forward processes with product reference states, the optimal reversal is the product of the individual optimal reversals.
  • In symmetric quantum hypothesis testing, the Fermi–Dirac thermal measurement has error exponent at most the classical Chernoff bound of its product-measurement-induced distributions, which is strictly below the quantum Chernoff exponent when the states do not commute; hence it cannot be 'pretty good' in the constant-factor sense.

Reading between the lines

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

  • If the positive-definiteness restriction can be lifted by a limiting argument, the same formula would govern optimal measurements for pure-state ensembles; a natural test is whether the closed form remains a minimizer when one $\sigma_x$ has a zero eigenvalue.
  • The duality with entropy-regularized semidefinite programs suggests that softmin thermal measurements may play a role in multiclass quantum machine learning analogous to thermal states in statistical mechanics, with the practical obstacle being the implementation of the operator exponential factor involving $A^*$.
  • The additivity property may yield a single-letter expression for the optimal error exponent in quantum Sanov-type settings with non-i.i.d. null hypotheses, since product structure of the optimal reversal should make large-deviation exponents additive.
  • The $\tau$-interpolation family suggests a one-parameter family of measurements whose hypothesis-testing performance interpolates between the constant-factor-guaranteed pretty good measurement at $\beta=1$ and the softmin thermal measurement at $\beta=0$; quantifying that trade-off is a direct next step.
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.

Referee Report

3 major / 3 minor

Summary. The paper develops a minimum change principle for quantum measurements based on quantum relative entropy. The forward process is a classical-to-quantum preparation channel and the reverse process a quantum-to-classical measurement channel. The main result (Theorem 1) gives a closed-form characterization of the optimal POVM as M*_x = τ^{-1/2} e^{ln σx - A*} τ^{-1/2}, where A* is the unique minimizer of a strictly convex dual functional. The framework is shown to recover the pretty good measurement and the Fermi–Dirac thermal measurement as special cases, and to yield a new family of softmin thermal measurements. The paper also proves an additivity property for the minimum change principle and analyzes the performance of Fermi–Dirac thermal measurements in quantum hypothesis testing, including non-asymptotic bounds and asymptotic error-exponent bounds.

Significance. The main theorem is a genuine and potentially useful unifying variational characterization. The proof is self-contained, including a direct-sum reduction, a careful derivation of the dual via minimax, and a strict convexity proof via an explicit Hessian computation. The recovery of known measurements as special cases of the reference state τ is conceptually appealing, and the new softmin thermal measurements come with an independent variational characterization in terms of entropy-regularized semidefinite programs. The additivity result and the hypothesis-testing bounds are nontrivial additions. However, the scope of the claims is currently limited by the full-rank assumption and by errors in two corollaries/proofs as discussed below.

major comments (3)
  1. [Section III.D, Eq. (51)] The expression for the Fermi–Dirac thermal measurement in Eq. (51) has the identity inside the exponential, M* = (e^{ln σ1 − ln σ0 + I})^{-1}. The proof in Appendix E (Eq. (E13)) and the subsequent thermal-state form in Eqs. (62)–(63) require the identity to be added outside the exponential, i.e., M* = (e^{ln σ1 − ln σ0} + I)^{-1}. As printed, Eq. (51) does not satisfy the completeness relation with Eq. (52), so this is a substantive error in a central corollary.
  2. [Appendix G, Eq. (G14)] The definition N^{(1)}_{x1} := Tr2[(I⊗τ2) ∑_{x2} M_{x1x2}] is inconsistent with the partial-trace computation in Eq. (G12), which requires N^{(1)}_{x1} = Tr2[(I⊗τ2^{1/2}) ∑_{x2} M_{x1x2} (I⊗τ2^{1/2})]. As written, the identification in Eq. (G13) of the reduced state with Qrev(τ1,N^{(1)}) is incorrect, and the same issue affects N^{(2)}_{x2} in Eq. (G23). The proof of Theorem 7 is therefore incomplete; it can be repaired by replacing τ2 with τ2^{1/2} in those definitions.
  3. [Theorem 1 / Corollary 3, full-rank assumptions] Theorem 1 and Corollary 3 assume σx>0 for all x, and Corollary 3 recovers the pretty good measurement only for full-rank substates. The standard pretty good measurement is most often applied to rank-deficient ensembles (e.g., pure states), where σ^{-1/2}σxσ^{-1/2} is well defined via the pseudoinverse when σ is full rank. The paper gives no limiting argument showing that the full-rank optimal POVM converges to the pseudoinverse PGM as σx approaches a rank-deficient limit, and the abstract's claim of recovering pretty good measurements is unqualified. This is a genuine gap in the claimed unification and should be addressed, either by stating the restriction explicitly or by supplying a rigorous perturbation argument.
minor comments (3)
  1. [Abstract] The abstract contains a spacing error: "asingleunconstrained" should read "a single unconstrained".
  2. [Section II.C.1, Eq. (27)] The gradient-descent update rule in Eq. (27) uses a fixed step size η∈(0,2); the local-convergence argument in Appendix C relies on the Hessian bound at A*, which is fine, but the statement that the algorithm has "guaranteed local convergence" should be made more precise about the neighborhood of A*.
  3. [Appendix G, Eq. (G18)] In the completeness check for N^{(1)}_{x1}, the text writes Tr2[(I⊗τ2)] = I; this is correct only because τ2 is a density operator, but the preceding definitions should be made consistent with the square-root factors as noted in the major comment.

Circularity Check

0 steps flagged · score 0.0 of 10

Self-contained convex derivation; known measurements are special cases, not assumed inputs.

full rationale

Theorem 1 is derived self-contained in Appendix B from the definition of Umegaki relative entropy, the direct-sum property, and a standard Lagrange/Slater minimax argument. No input parameter is fitted: the dual variable A arises as a Lagrange multiplier, the stationarity condition sum_x e^{ln sigma_x - A*} = tau is derived rather than imposed, and the optimal POVM in (25) follows by substitution. The special cases are genuine specializations: Corollary 3 sets tau = sigma and obtains the pretty good measurement, Corollary 4 sets tau = I/d, and Corollary 5 further specializes to the binary case and derives the Fermi-Dirac form from the first-order stationarity condition. None of these steps assumes the target measurement as an input. The proof of Theorem 6 is likewise self-contained and does not depend on Theorem 1 except for motivation. Self-citations to [16] and [26] appear in contextual statements such as 'already observed in [16]' and 'extending our earlier developments in [26]', but they are not load-bearing: every claimed recovery is established by the paper's own equations. The full-rank assumptions tau > 0 and sigma_x > 0 are explicit hypotheses, not hidden inputs; the absence of a limiting argument for rank-deficient ensembles is a scope limitation, not circularity. Therefore no circular step is present.

Assumptions & free parameters 2 free parameters · 7 assumptions · 0 invented entities

The central claim rests on standard convex analysis and quantum relative entropy identities, plus explicit full-rank assumptions. The only tunable input is the reference state tau, which parametrizes the measurement family, and the temperature T in Theorem 6; no numerical data are fitted. The paper contributes the optimization principle and the new softmin thermal family, not new physical entities.

free parameters (2)
  • Reference state tau = not fitted; arbitrary positive definite density operator
    Theorem 1 defines the optimal measurement family only after choosing tau; special choices sigma, I/d, and e^{-beta H}/Z recover PGM, softmin thermal, and interpolated measurements.
  • Temperature T in entropy-regularized SDP = T > 0 fixed by user
    Theorem 6 introduces T as a regularization strength; it is not fitted to data.
assumptions (7)
  • domain assumption Finite-dimensional Hilbert space and finite alphabet X
    The trace, matrix exponential, and POVM formalism in Section II are finite-dimensional; d in N and r = |X| in N.
  • domain assumption tau is positive definite and normalized
    Theorem 1 requires tau > 0 so that tau^{-1/2} exists and Tr[tau] = 1 for the dual identity in (B22).
  • domain assumption Each sigma_x = p(x) rho_x is positive definite
    Theorem 1 requires sigma_x > 0 so that ln sigma_x and e^{ln sigma_x - A} are defined.
  • domain assumption The reverse process is restricted to quantum-to-classical measurement channels of the form in Eq. (13)
    The theorem solves the minimum change principle over POVMs, not over arbitrary quantum channels; this is the paper's chosen setting, stated in Section IIB.
  • standard math Direct-sum additivity, tensor additivity, and superadditivity of Umegaki relative entropy
    Used in (B1) for the direct-sum decomposition and in (G3) and (G5) for the additivity proof.
  • standard math Slater's condition and minimax equality for the convex dual
    Used to swap min and sup in (B14) and (F4); the feasible set has interior points N_x = tau/|X| or M_x = I/|X|.
  • standard math Araki-Lieb-Thirring inequality with equality conditions
    Used in Lemma 17 to prove strict inequality of measured Renyi divergences for noncommuting states.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unifying quantum measurement constructions via a relative-entropy minimum change principle." pith.science (2026). https://pith.science/paper/USCEUV5C

@misc{pith2026260804055,
  author       = {Pith},
  title        = {Pith review of: Unifying quantum measurement constructions via a relative-entropy minimum change principle},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/USCEUV5C}},
  note         = {Machine review of arXiv:2608.04055}
}
read the original abstract

The minimum change principle provides an information-theoretic characterization of the Bayes reversal channel in classical probability theory and has recently been proposed as a framework for extending Bayes' rule to quantum information theory. Using quantum relative entropy, we investigate a minimum change principle for the setting of quantum statistical inference. Specifically, we consider a forward process based on a classical-to-quantum preparation channel and a reverse process based on a quantum-to-classical measurement channel. We establish a closed-form characterization of measurements that are optimal for this principle, and this optimal measurement can be found via a dual formulation involving a single unconstrained Hermitian variable. This perspective allows us to recover some notable measurements within the same framework, including pretty good measurements and Fermi-Dirac thermal measurements, and we use it to discover a novel family that we call softmin thermal measurements. We further show that softmin thermal measurements arise as optimal solutions to entropy-regularized semidefinite optimization problems, demonstrating that they play a role for measurements analogous to that of thermal states in statistical mechanics. Finally, we prove an additivity property for the relative-entropy minimum change principle and investigate the performance of Fermi-Dirac thermal measurements for quantum hypothesis testing.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

109 extracted references · 52 canonical work pages

  1. [1]

    The value in doing so is that, once one has the optimalA ⋆, the optimal measurement in (25) can be constructed directly

    Computing the optimal measurement in Theorem 1 Here we devise an algorithm for finding the optimal value of the dual objective in (23). The value in doing so is that, once one has the optimalA ⋆, the optimal measurement in (25) can be constructed directly. Given that the function in (26) is strictly convex inA, one can employ gradient descent in order to ...

  2. [2]

    exponential family

    Binary classical case Here we briefly elaborate on the form of the Bayesian reversal channel when the input alphabetX is binary. This further specializes the classical case mentioned above and prepares for the developments in Section IVB2. Suppose that there is a binary random variableXwith a realizationx∈ {0,1}and probabilityp∈(0,1)for x= 0and probabilit...

  3. [3]

    Exponential families Suppose that there is a prior probability distribution pX (x)for the forward process, wherex∈ X. Suppose nowthattheforwardchannelisintheexponentialfamily: pY|X (y|x) = e−Hx(y) Zx ,(65) Zx := X y∈Y e−Hx(y),(66) whereH x(y)is a Hamiltonian function forx∈ X(i.e., eachH x is a polynomial function of the elements iny). 8 Then, after defini...

  4. [4]

    In this case, the prior probability densities in (65) are known as Boltzmann machines [54, 55]

    Sigmoid neurons and Boltzmann machines Suppose furthermore that each Hamiltonian function is quadratic, so that Hx(y) =y T Wxy+w T x y,(76) whereW x ∈R n×n andw x ∈R n forx∈ {0,1}. In this case, the prior probability densities in (65) are known as Boltzmann machines [54, 55]. Then an immediate conclusion from (71)–(72) is that pX|Y (0|y) = 1 1 +e yT W y+w...

  5. [5]

    pretty good,

    Review of quantum hypothesis testing Let us begin by reviewing some well known results in quantum hypothesis testing. In the scenario of binary symmetric hypothesis testing, a stateρ0 is prepared with probabilityp 0 and a stateρ1 is prepared with probability p1, wherep 0 +p 1 = 1. For simplicity, we assume throughout thatρ 0 andρ 1 are positive definite s...

  6. [6]

    0” and the second with guessing “1

    Non-asymptotic performance of Fermi–Dirac thermal measurements We now establish upper bounds on the error probability in hypothesis testing when using Fermi–Dirac thermal measurements. Defining A := lnσ 1 −lnσ 0,(108) we can write the Fermi–Dirac thermal measurement in (51) as follows: eA +I −1 , e−A +I −1 ,(109) where the first outcome is identified with...

  7. [7]

    preparation

    Asymptotic performance of Fermi–Dirac thermal measurements In this section, we discuss the performance of Fermi– Dirac thermal measurements in asymptotic hypothesis testing and establish an upper bound on the error probability when doing so. We also mention how Fermi– Dirac thermal measurements for i.i.d. states can be realized as a product measurement fo...

  8. [8]

    E. T. Jaynes,Probability Theory: The Logic of Science, edited by G. L. Bretthorst (Cambridge University Press, Cambridge, 2003)

    2003

Show all 109 references
  1. [9]

    Schack, T

    R. Schack, T. A. Brun, and C. M. Caves, Quantum Bayes rule, Physical Review A64, 014305 (2001)

    2001

  2. [10]

    M. S. Leifer and R. W. Spekkens, Towards a formulation of quantum theory as a causally neutral theory of Bayesianinference,PhysicalReviewA88,052130(2013)

    2013

  3. [11]

    A. J. Parzygnat and B. P. Russo, A non-commutative Bayes’ theorem, Linear Algebra and Its Applications 644, 28 (2022)

    2022

  4. [12]

    A. J. Parzygnat and J. Fullwood, From time-reversal symmetry to quantum Bayes’ rules, PRX Quantum4, 020334 (2023)

    2023

  5. [13]

    A. C. Cenxin, K. Onggadinata, D. Kaszlikowski, and V. Scarani, Quantum Bayesian inference in quasiprobability representations, PRX Quantum4, 020352 (2023)

    2023

  6. [14]

    Belavkin, Optimal distinction of non-orthogonal quantum signals, Radio Engineering and Electronic Physics20, 39 (1975)

    V. Belavkin, Optimal distinction of non-orthogonal quantum signals, Radio Engineering and Electronic Physics20, 39 (1975)

    1975

  7. [15]

    Belavkin, Optimal multiple quantum statistical hypothesis testing, Stochastics1, 315 (1975)

    V. Belavkin, Optimal multiple quantum statistical hypothesis testing, Stochastics1, 315 (1975)

    1975

  8. [16]

    L. P. Hughston, R. Jozsa, and W. K. Wootters, A complete classification of quantum ensembles having a given density matrix, Physics Letters A183, 14 (1993)

    1993

  9. [17]

    Hausladen and W

    P. Hausladen and W. K. Wootters, A ‘pretty good’ measurement for distinguishing quantum states, Journal of Modern Optics41, 2385 (1994)

    1994

  10. [18]

    Petz, Sufficient subalgebras and the relative entropy 14 of states of a von Neumann algebra, Communications in Mathematical Physics105, 123 (1986)

    D. Petz, Sufficient subalgebras and the relative entropy 14 of states of a von Neumann algebra, Communications in Mathematical Physics105, 123 (1986)

    1986

  11. [19]

    Petz, Sufficiency of channels over von Neumann algebras, The Quarterly Journal of Mathematics39, 97 (1988)

    D. Petz, Sufficiency of channels over von Neumann algebras, The Quarterly Journal of Mathematics39, 97 (1988)

    1988

  12. [20]

    Barnum and E

    H. Barnum and E. Knill, Reversing quantum dynamics with near-optimal quantum and classical fidelity, Journal of Mathematical Physics43, 2097 (2002)

    2002

  13. [21]

    Hayden, R

    P. Hayden, R. Jozsa, D. Petz, and A. Winter, Structure of states which satisfy strong subadditivity of quantum entropy with equality, Communications in Mathematical Physics246, 359 (2004)

    2004

  14. [22]

    Junge, R

    M. Junge, R. Renner, D. Sutter, M. M. Wilde, and A. Winter, Universal recovery maps and approximate sufficiency of quantum relative entropy, Annales Henri Poincaré19, 2955 (2018)

    2018

  15. [23]

    G. Bai, F. Buscemi, and V. Scarani, Quantum Bayes’ rule and Petz transpose map from the minimum change principle, Physical Review Letters135, 090203 (2025)

    2025

  16. [24]

    Aitchison, Goodness of prediction fit, Biometrika62, 547 (1975)

    J. Aitchison, Goodness of prediction fit, Biometrika62, 547 (1975)

    1975

  17. [25]

    May and W

    S. May and W. L. Harper, Toward an optimization procedure for applying minimum change principles in probability kinematic, inFoundations of Probability Theory, Statistical Inference, and Statistical Theories of Science. Volume I: Foundations and Philosophy of Epistemic Applica...

    1976

  18. [26]

    P. M. Williams, Bayesian conditionalisation and the principle of minimum information, The British Journal for the Philosophy of Science31, 131 (1980)

    1980

  19. [27]

    Zellner, Optimal information processing and Bayes’s theorem, The American Statistician42, 278 (1988)

    A. Zellner, Optimal information processing and Bayes’s theorem, The American Statistician42, 278 (1988)

    1988

  20. [28]

    transition probability

    A. Uhlmann, The “transition probability” in the state space of a ∗-algebra, Reports on Mathematical Physics 9, 273 (1976)

    1976

  21. [29]

    C. W. Helstrom,Quantum Detection and Estimation Theory(Academic Press, New York, 1976)

    1976

  22. [30]

    A. S. Holevo,Probabilistic and Statistical Aspects of Quantum Theory, 2nd ed. (Edizioni della Normale, Pisa, 2011)

    2011

  23. [31]

    Hayashi, ed.,Asymptotic Theory of Quantum Statistical Inference: Selected Papers(World Scientific, Hackensack, NJ, 2005)

    M. Hayashi, ed.,Asymptotic Theory of Quantum Statistical Inference: Selected Papers(World Scientific, Hackensack, NJ, 2005)

    2005

  24. [32]

    Umegaki, Conditional expectation in an operator algebra, IV (entropy and information), inKodai Mathematical Seminar Reports, Vol

    H. Umegaki, Conditional expectation in an operator algebra, IV (entropy and information), inKodai Mathematical Seminar Reports, Vol. 14 (Department of Mathematics, Tokyo Institute of Technology, 1962) pp. 59–85

    1962

  25. [33]

    Liu and M

    N. Liu and M. M. Wilde, Fermi-Dirac thermal measurements: A framework for quantum hypothesis testing and semidefinite optimization (2026), arXiv:2603.04061 [quant-ph]

    2026

  26. [34]

    A. He, N. Liu, and M. M. Wilde, Fermi-dirac machines as quantizations of neurons (2026), arXiv:2605.24386 [quant-ph]

    2026 arXiv

  27. [35]

    A. He, N. Liu, and M. M. Wilde, Canonical quantization of neurons (2026), arXiv:2607.05000 [quant-ph]

    2026 arXiv

  28. [36]

    Lindsey, Fast randomized entropically regularized semidefinite programming (2023), arXiv:2303.12133 [math.OC]

    M. Lindsey, Fast randomized entropically regularized semidefinite programming (2023), arXiv:2303.12133 [math.OC]

    2023 arXiv

  29. [37]

    J. S. Bridle, Probabilistic interpretation of feedforward classification network outputs, with relationships to statistical pattern recognition, inNeurocomputing: Algorithms, Architectures and Applications, NATO ASI Series F: Computer and Systems Sciences, Vol. 68, edited by F...

    1990

  30. [38]

    C.M.Bishop,Pattern Recognition and Machine Learning (Springer, New York, 2006)

    2006

  31. [39]

    Goodfellow, Y

    I. Goodfellow, Y. Bengio, and A. Courville,Deep Learning(MIT Press, Cambridge, MA, 2016)

    2016

  32. [40]

    A. S. Holevo, Statistical decision theory for quantum systems, Journal of Multivariate Analysis3, 337 (1973)

    1973

  33. [41]

    E. T. Jaynes, Information theory and statistical mechanics, Physical Review106, 620 (1957)

    1957

  34. [42]

    E. T. Jaynes, Information theory and statistical mechanics. II, Physical Review108, 171 (1957)

    1957

  35. [43]

    E. T. Jaynes, Information theory and statistical mechanics (1962) reprinted in E. T. Jaynes: Papers on Probability, Statistics and Statistical Physics, R.D. Rosenkrantz (ed.), D. Reidel Publishing Company, 1983

    1962

  36. [44]

    Csiszár,I-Divergence Geometry of Probability Distributions and Minimization Problems, The Annals of Probability3, 146 (1975)

    I. Csiszár,I-Divergence Geometry of Probability Distributions and Minimization Problems, The Annals of Probability3, 146 (1975)

    1975

  37. [45]

    Csiszár, Sanov property, generalizedI-projection and a conditional limit theorem, The Annals of Probability 12, 768 (1984)

    I. Csiszár, Sanov property, generalizedI-projection and a conditional limit theorem, The Annals of Probability 12, 768 (1984)

    1984

  38. [46]

    Csiszár and F

    I. Csiszár and F. Matúš, Information projections revisited, IEEE Transactions on Information Theory49, 1474 (2003)

    2003

  39. [47]

    Nuradha, H

    T. Nuradha, H. K. Mishra, F. Leditzky, and M. M. Wilde, Multivariate fidelities, Journal of Physics A: Mathematical and Theoretical58, 165304 (2025)

    2025

  40. [48]

    Girardi, A

    F. Girardi, A. Oufkir, B. Regula, M. Tomamichel, M. Berta, and L. Lami, Quantum umlaut information (2025), arXiv:2503.21479 [quant-ph]

    2025

  41. [49]

    Y. Qiu, K. Koor, and P. Rebentrost, The quantum Esscher transform (2025), arXiv:2401.07561 [quant-ph]

    2025 arXiv

  42. [50]

    L. Lami, M. Berta, and B. Regula, Asymptotic quantification of entanglement with a single copy, Nature Physics22, 439 (2026)

    2026

  43. [51]

    Nocedal and S

    J. Nocedal and S. J. Wright,Numerical Optimization, 2nd ed., Springer Series in Operations Research and Financial Engineering (Springer, New York, NY, 2006)

    2006

  44. [52]

    A. M. Alhambra, Quantum many-body systems in thermal equilibrium, PRX Quantum4, 040201 (2023)

    2023

  45. [53]

    F. G. S. L. Brandao and K. M. Svore, Quantum speed- ups for solving semidefinite programs, in2017 IEEE 58th Annual Symposium on Foundations of Computer Science (FOCS)(2017) pp. 415–426

    2017

  46. [54]

    F. G. S. L. Brandão, A. Kalev, T. Li, C. Y.-Y. Lin, K. M. Svore, and X. Wu, Quantum SDP Solvers: Large Speed-Ups, Optimality, and Applications to Quantum Learning,in46th International Colloquium on Automata, Languages, and Programming (ICALP 2019), Leibniz International Procee...

    2019

  47. [55]

    van Apeldoorn and A

    J. van Apeldoorn and A. Gilyén, Improvements in Quantum SDP-Solving with Applications, in46th International Colloquium on Automata, Languages, and Programming (ICALP 2019), Leibniz International Proceedings in Informatics (LIPIcs), Vol. 132, edited by C. Baier, I. Chatzigianna...

    2019

  48. [56]

    van Apeldoorn, A

    J. van Apeldoorn, A. Gilyén, S. Gribling, and R. de Wolf, Quantum SDP-Solvers: Better upper and lower bounds, Quantum4, 230 (2020)

    2020

  49. [57]

    N. Liu, M. Minervini, D. Patel, and M. M. Wilde, Quantum thermodynamics and semi-definite optimization (2025), arXiv:2505.04514 [quant-ph]

    2025 arXiv

  50. [58]

    M. H. Amin, E. Andriyash, J. Rolfe, B. Kulchytskyy, and R. Melko, Quantum Boltzmann machine, Physical Review X8, 021050 (2018)

    2018

  51. [59]

    Kieferová and N

    M. Kieferová and N. Wiebe, Tomography and generative training with quantum Boltzmann machines, Physical Review A96, 062327 (2017)

    2017

  52. [60]

    Benedetti, J

    M. Benedetti, J. Realpe-Gómez, R. Biswas, and A. Perdomo-Ortiz, Quantum-assisted learning of hardware-embedded probabilistic graphical models, Physical Review X7, 041052 (2017)

    2017

  53. [61]

    D. H. Ackley, G. E. Hinton, and T. J. Sejnowski, A learning algorithm for Boltzmann machines, Cognitive Science9, 147 (1985)

    1985

  54. [62]

    G. E. Hinton and T. J. Sejnowski, Learning and relearning in Boltzmann machines, inParallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1, edited by D. E. Rumelhart and J. L. McClelland (MIT Press, 1986) pp. 282–317

    1986

  55. [63]

    Boyd and L

    S. Boyd and L. Vandenberghe,Convex Optimization (Cambridge University Press, 2004)

    2004

  56. [64]

    C. W. Helstrom, Detection theory and quantum mechanics, Information and Control10, 254 (1967)

    1967

  57. [65]

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

    1969

  58. [66]

    A. S. Holevo, The analogue of statistical decision theory in the noncommutative probability theory, Trudy Moskovskogo Matematicheskogo Obshchestva (Proceedings of the Moscow Mathematical Society)26, 133 (1972)

    1972

  59. [67]

    Watrous,The Theory of Quantum Information (Cambridge University Press, Cambridge, 2018)

    J. Watrous,The Theory of Quantum Information (Cambridge University Press, Cambridge, 2018)

    2018

  60. [68]

    H. K. Mishra, L. Lami, and M. M. Wilde, Near- optimal performance of square-root measurement for general score functions and quantum ensembles (2025), arXiv:2505.20447 [quant-ph]

    2025 arXiv

  61. [69]

    K. M. R. Audenaert, J. Calsamiglia, R. Muñoz-Tapia, E. Bagan, L. Masanes, A. Acin, and F. Verstraete, Discriminating states: The quantum Chernoff bound, Physical Review Letters98, 160501 (2007)

    2007

  62. [70]

    Nussbaum and A

    M. Nussbaum and A. Szkoła, The Chernoff lower bound for symmetric quantum hypothesis testing, Annals of Statistics37, 1040 (2009)

    2009

  63. [71]

    Chernoff, A measure of asymptotic efficiency for tests of a hypothesis based on the sum of observations, The Annals of Mathematical Statistics23, 493 (1952)

    H. Chernoff, A measure of asymptotic efficiency for tests of a hypothesis based on the sum of observations, The Annals of Mathematical Statistics23, 493 (1952)

    1952

  64. [72]

    K. Ji, H. K. Mishra, M. Mosonyi, and M. M. Wilde, Barycentric bounds on the error exponents of quantum hypothesis exclusion, IEEE Transactions on Information Theory72, 2391 (2026)

    2026

  65. [73]

    Bjelaković, J.-D

    I. Bjelaković, J.-D. Deuschel, T. Krüger, R. Seiler, R. Siegmund-Schultze, and A. Szkoła, A quantum version of sanov’s theorem, Communications in Mathematical Physics260, 659 (2005)

    2005

  66. [74]

    Bjelaković, J.-D

    I. Bjelaković, J.-D. Deuschel, T. Krüger, R. Seiler, R. Siegmund-Schultze, and A. Szkoła, Typical support and sanov large deviations of correlated states, Communications in Mathematical Physics279, 559 (2008)

    2008

  67. [75]

    J. Nötzel, Hypothesis testing on invariant subspaces of the symmetric group, part i: Quantum sanov’s theorem and arbitrarily varying sources, Journal of Physics A: Mathematical and Theoretical47, 235303 (2014)

    2014

  68. [76]

    Lami, Generalised quantum sanov theorem revisited (2025), arXiv:2510.06340 [quant-ph]

    L. Lami, Generalised quantum sanov theorem revisited (2025), arXiv:2510.06340 [quant-ph]

    2025

  69. [77]

    Meunson and T

    A. Meunson and T. Deesuwan, Cumulant-based quantum relative Rényi functional (2026), arXiv:2606.31205 [quant-ph]

    2026 arXiv

  70. [78]

    F.HiaiandM.Mosonyi,Differentquantumf-divergences and the reversibility of quantum operations, Reviews in Mathematical Physics29, 1750023 (2017)

    2017

  71. [79]

    Hiai and M

    F. Hiai and M. Mosonyi, Test-measured Rényi divergences, IEEE Transactions on Information Theory 69, 1074 (2023)

    2023

  72. [80]

    Fawzi and J

    H. Fawzi and J. Saunderson, Optimal self-concordant barriers for quantum relative entropies, SIAM Journal on Optimization33, 2858 (2023)

    2023

  73. [81]

    K. He, J. Saunderson, and H. Fawzi, Interior point methods for structured quantum relative entropy optimization problems (2025), arXiv:2407.00241 [quant- ph]

    2025 arXiv

  74. [82]

    K. He, J. Saunderson, and H. Fawzi, QICS: Quantum information conic solver (2025), arXiv:2410.17803 [math.OC]

    2025 arXiv

  75. [83]

    Koßmann and R

    G. Koßmann and R. Schwonnek, Optimising the relative entropy under semidefinite constraints, npj Quantum Information12, 23 (2026)

    2026

  76. [84]

    Sreekumar, Z

    S. Sreekumar, Z. Goldfeld, and M. M. Wilde, Performance guarantees for quantum neural estimation of entropies, Quantum10, 2113 (2026)

    2026

  77. [85]

    Falk, Inequalities of J

    H. Falk, Inequalities of J. W. Gibbs, American Journal of Physics38, 858 (1970)

    1970

  78. [86]

    M. M. Wilde, Quantum Fisher information matrices from Rényi relative entropies (2025), arXiv:2510.02218 [quant- ph]

    2025

  79. [87]

    Beigi, Sandwiched Rényi divergence satisfies data processing inequality, Journal of Mathematical Physics 54, 122202 (2013)

    S. Beigi, Sandwiched Rényi divergence satisfies data processing inequality, Journal of Mathematical Physics 54, 122202 (2013)

    2013

  80. [88]

    Nesterov,Lectures on Convex Optimization, 2nd ed., Springer Optimization and Its Applications, Vol

    Y. Nesterov,Lectures on Convex Optimization, 2nd ed., Springer Optimization and Its Applications, Vol. 137 (Springer, Cham, 2018)

    2018

  81. [89]

    Bubeck, Convex optimization: Algorithms and complexity,FoundationsandTrendsinMachineLearning 8, 231 (2015)

    S. Bubeck, Convex optimization: Algorithms and complexity,FoundationsandTrendsinMachineLearning 8, 231 (2015)

    2015

  82. [90]

    E. A. Carlen, Trace inequalities and quantum entropy: An introductory course, inEntropy and the Quantum, Contemporary Mathematics, Vol. 529 (American Mathematical Society, Providence, RI, 2010) pp. 73–140

    2010

  83. [91]

    Capel, A

    A. Capel, A. Lucia, and D. Perez-Garcia, Superadditivity of quantum relative entropy for general states, IEEE Transactions on Information Theory64, 4758 (2018)

    2018

  84. [92]

    Berta, O

    M. Berta, O. Fawzi, and M. Tomamichel, On variational expressions for quantum relative entropies, Letters in Mathematical Physics107, 2239 (2017), 1512.02615v2

    2017 arXiv

  85. [93]

    Petz, Quasi-entropies for states of a von Neumann algebra, Publications of the Research Institute for Mathematical Sciences21, 787 (1985)

    D. Petz, Quasi-entropies for states of a von Neumann algebra, Publications of the Research Institute for Mathematical Sciences21, 787 (1985)

    1985

  86. [94]

    Petz, Quasi-entropies for finite quantum systems, Reports on Mathematical Physics23, 57 (1986)

    D. Petz, Quasi-entropies for finite quantum systems, Reports on Mathematical Physics23, 57 (1986)

    1986

  87. [95]

    Müller-Lennert, F

    M. Müller-Lennert, F. Dupuis, O. Szehr, S. Fehr, and M. Tomamichel, On quantum Rényi entropies: A new generalization and some properties, Journal of 16 Mathematical Physics54, 122203 (2013), 1306.3142

    2013 arXiv

  88. [96]

    M. M. Wilde, A. Winter, and D. Yang, Strong converse for the classical capacity of entanglement-breaking and Hadamard channels via a sandwiched Rényi relative entropy, Communications in Mathematical Physics331, 593 (2014), 1306.1586

    2014 arXiv

  89. [97]

    Fuchs,Distinguishability and Accessible Information in Quantum Theory, Ph.D

    C. Fuchs,Distinguishability and Accessible Information in Quantum Theory, Ph.D. thesis, University of New Mexico (1996), arXiv:quant-ph/9601020

    1996 arXiv

  90. [98]

    R. L. Frank and E. H. Lieb, Monotonicity of a relative Rényi entropy, Journal of Mathematical Physics54, 122201 (2013)

    2013

  91. [99]

    Datta and F

    N. Datta and F. Leditzky, A limit of the quantum Rényi divergence, Journal of Physics A: Mathematical and Theoretical47, 045304 (2014)

    2014

  92. [100]

    ∞X n=0 (lnσ x −A) n n! # (B28) = ∂ ∂aji Tr

    F. Hiai, Equality cases in matrix norm inequalities of Golden-Thompson type, Linear and Multilinear Algebra 36, 239 (1994). Appendix A: Alternative minimum change principle in Equation(20) In the main text, we focused on the minimum change principle based on minimizing D(Qrev(...

    1994

  93. [101]

    Proof of Equation(111) Consider that e−A +I −1 = 1 2 (I+ tanh(A/2)),(H1) eA +I −1 =I− e−A +I −1 (H2) =I− 1 2 (I+ tanh(A/2))(H3) = 1 2 (I−tanh(A/2)),(H4) implying that Tr h e−A +I −1 σ0 i + Tr h eA +I −1 σ1 i = Tr 1 2 (I+ tanh(A/2))σ 0 + Tr 1 2 (I−tanh(A/2))σ 1 (H5) = 1 2 − 1 2...

  94. [102]

    Supplementary lemma Lemma 15.The following inequality holds for alls∈[0,1]: (1 +e x)−1 ≤e −h(s)e−sx,(H8) where the binary entropy is defined ash(s):=−slns−(1−s) ln(1−s). Proof.We would like the following inequality to hold for allx∈R, for some functionf(s): (1 +e x)−1 ≤f(s)e −...

  95. [103]

    Now pickα=s

    Proof of Lemma 9 Consider that, forα∈(0,1), by applying [85, Lemma 3], Tr ρα 0 ρ1−α 1 ≤inf ω>0 (Tr[ωρ0])α Tr h ω α α−1 ρ1 i 1−α (H31) ≤ Tr e(1−s)(lnρ 1−lnρ 0)ρ0 α Tr h e(1−s)(lnρ 1−lnρ 0) α α−1 ρ1 i 1−α ,(H32) where the last inequality follows by pickingω=e(1−s)(lnρ 1−lnρ 0). ...

  96. [104]

    Proof of Lemma 8 Consider that Tr h e−A +I −1 σ0 i + Tr h eA +I −1 σ1 i ≤s s (1−s) 1−s Tr e(1−s)Aσ0 +s s (1−s) 1−s Tr e−sAσ1 (H50) =s s (1−s) 1−s Tr e(1−s)(lnσ 1−lnσ 0)σ0 + Tr e−s(lnσ 1−lnσ 0)σ1 (H51) =s s (1−s) 1−s ps 0p1−s 1 Tr e(1−s)(lnρ 1−lnρ 0)ρ0 + Tr e−s(lnρ 1−lnρ 0)ρ1 (...

  97. [105]

    Proof of Proposition 10 Proof of Proposition 10.Consider that Tr e−A(n) +I −1 p0ρ⊗n 0 + Tr eA(n) +I −1 p1ρ⊗n 1 ≤2g(s, p0)e−min {C♮ s(ρ⊗n 0 ∥ρ⊗n 1 ),C♮ 1−s(ρ⊗n 1 ∥ρ⊗n 0 )} (H56) = 2g(s, p0)e−min {nC♮ s(ρ0∥ρ1),nC♮ 1−s(ρ1∥ρ0)} (H57) = 2g(s, p0)e−nmin {C♮ s(ρ0∥ρ1),C♮ 1−s(ρ1∥ρ0)},(...

  98. [106]

    states Here we elaborate on a method for realizing the Fermi–Dirac thermal measurement in (124) and (123) by means of a product measurement followed by classical postprocessing

    Implementing Fermi–Dirac thermal measurement for i.i.d. states Here we elaborate on a method for realizing the Fermi–Dirac thermal measurement in (124) and (123) by means of a product measurement followed by classical postprocessing. Set H := lnρ 1 −lnρ 0,(H66) and let a spect...

  99. [107]

    Fori∈[n], measure theith system in the eigenbasis{|ϕ j⟩}j ofHand record the outcome as hji

  100. [108]

    Computea(j n)≡ln p1 p0 + Pn i=1 hji

  101. [109]

    0” (i.e., the state isρ0) with probability ea(jn) + 1 −1 , and output “1

    Output “0” (i.e., the state isρ0) with probability ea(jn) + 1 −1 , and output “1” (i.e., the state isρ 1) with probability e−a(jn) + 1 −1 . The first step above implements a product measurement, and the last two steps realize classical postprocessing. Underthisscheme, foranarb...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.