Quantum Fisher information matrix via its classical counterpart from random measurements

Jianfeng Lu; Kecen Sha

arxiv: 2509.08196 · v4 · submitted 2025-09-10 · 🪐 quant-ph · math-ph· math.MP· math.OC

Quantum Fisher information matrix via its classical counterpart from random measurements

Jianfeng Lu , Kecen Sha This is my paper

Pith reviewed 2026-05-18 18:41 UTC · model grok-4.3

classification 🪐 quant-ph math-phmath.MPmath.OC

keywords quantum Fisher informationclassical Fisher informationHaar random measurementsquantum metrologyconcentration boundsvariational quantum algorithmsrandomized measurements

0 comments

The pith

Averaging classical Fisher information over Haar-random bases equals half the quantum Fisher information matrix for pure states.

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

The paper establishes that the expected value of the classical Fisher information matrix, when averaged over Haar-random measurement bases, exactly equals one-half of the quantum Fisher information matrix for pure states. This identity is reviewed through its link to covariant measurements in quantum metrology. The authors further compute the exact variance of the averaged matrix, which scales as O(N^{-1}), and derive non-asymptotic concentration bounds of the form exp(-Θ(N)t²). These bounds show that only a small number of random bases are needed to approximate the quantum Fisher information matrix with high accuracy in high-dimensional spaces. The results supply a theoretical foundation for using randomized measurements to precondition variational quantum algorithms more efficiently.

Core claim

Averaging the classical Fisher information matrix over Haar-random measurement bases yields E_{U~μ_H}[F^U(θ)] = 1/2 Q(θ) for pure states in C^N, with exact variance O(N^{-1}) and non-asymptotic concentration bounds exp(-Θ(N)t²) showing few samples suffice for accurate approximation.

What carries the argument

The averaging of the classical Fisher information matrix over the Haar measure on unitary bases, which produces the exact factor-of-1/2 relation to the quantum Fisher information matrix.

If this is right

Few random measurement bases suffice to approximate the quantum Fisher information matrix accurately in high dimensions.
The approach connects classical and quantum Fisher information through covariant measurements.
The concentration bounds guarantee reliable approximation even with finite numbers of random bases.
This supplies a practical route to preconditioning in quantum variational algorithms without full quantum Fisher information computation.

Where Pith is reading between the lines

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

The method could lower the state-preparation overhead in variational quantum eigensolvers that rely on natural gradients.
Similar averaging arguments might extend to other information measures or to approximate quantum geometric tensors.
The exponential concentration in dimension suggests robustness on near-term devices where only modest numbers of measurements are feasible.

Load-bearing premise

The averaging identity and the variance and concentration bounds are stated specifically for pure states.

What would settle it

A direct numerical check for a known pure state in dimension N=8 where the sample-averaged classical Fisher matrix deviates from half the quantum Fisher matrix by more than the predicted O(N^{-1}) variance would falsify the central identity.

read the original abstract

Preconditioning with the quantum Fisher information matrix (QFIM) is a popular approach in quantum variational algorithms. Yet the QFIM is costly to obtain directly, usually requiring more state preparation than its classical counterpart: the classical Fisher information matrix (CFIM). It is known that averaging the classical Fisher information matrix over Haar-random measurement bases yields $\mathbb{E}_{U\sim\mu_H}[F^U(\boldsymbol{\theta})] = \frac{1}{2}Q(\boldsymbol{\theta})$ for pure states in $\mathbb{C}^N$. In this paper, we review this identity by revealing its connection to covariant measurement in quantum metrology. Furthermore, we go beyond this and obtain the exact variance of CFIM ($O(N^{-1})$), estimate its moment, and establish non-asymptotic concentration bounds ($\exp(-\Theta(N)t^2)$), demonstrating that using few random measurement bases is sufficient to approximate the QFIM accurately in high-dimensional settings. This work establishes a solid theoretical foundation for efficient quantum natural gradient methods via randomized measurements.

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.

Desk Editor's Note private letter to a colleague

The paper strengthens a known averaging identity with explicit variance O(N^{-1}) and concentration bounds exp(-Θ(N)t²) for approximating QFIM from random CFIMs on pure states.

read the letter

The main thing to know is that averaging classical Fisher information matrices over Haar-random bases recovers half the quantum Fisher information matrix for pure states, and the new material is the exact variance scaling plus non-asymptotic concentration that shows a small number of samples already gives a reliable estimate in high dimensions. The concentration bound looks practically relevant for cutting the cost of preconditioning in variational quantum algorithms. The paper reviews the identity through its link to covariant measurements and then derives the variance and tail bounds from that starting point. Those derivations appear independent of the earlier literature they cite, which is a plus. The variance result is clean once the expectation identity is in hand, and the exponential tail bound supports the claim that few random measurements suffice. The main limitation is the explicit restriction to pure states in C^N; nothing in the work extends the bounds to mixed states, so readers working with density matrices will still need the direct QFIM route. The abstract states the assumptions clearly, but the full derivations would be needed to check whether any extra conditions on the parameterization slipped in. This is useful for people already using quantum natural gradient methods or randomized metrology protocols who want a cheaper surrogate for the QFIM. A reader focused on resource-efficient variational algorithms would get concrete guidance from the concentration analysis. The work is grounded enough and the new bounds are sharp enough that it should go to peer review rather than a desk reject.

Referee Report

0 major / 4 minor

Summary. The paper reviews the known identity that averaging the classical Fisher information matrix (CFIM) F^U(θ) over Haar-random measurement bases U yields E_{U~μ_H}[F^U(θ)] = (1/2) Q(θ) for pure states in C^N. It connects this identity to covariant measurements in quantum metrology, derives the exact variance of the CFIM (scaling as O(N^{-1})), estimates moments, and establishes non-asymptotic concentration bounds of the form exp(-Θ(N) t²), concluding that few random bases suffice to approximate the QFIM accurately for preconditioning in quantum variational algorithms.

Significance. If the derivations hold, the work supplies concrete non-asymptotic guarantees on the sample complexity of randomized CFIM averaging, which directly supports efficient quantum natural gradient methods by reducing the state-preparation overhead relative to direct QFIM estimation. The exact variance scaling and matrix concentration results are particularly useful for high-dimensional settings, and the link to covariant measurements provides a useful bridge to quantum metrology. These elements together offer a solid theoretical foundation for the proposed approximation technique.

minor comments (4)

[Abstract] Abstract: the phrase 'estimate its moment' is vague; the main text should specify which moment (e.g., second or higher) is computed and how it relates to the variance result.
[Review section] Review section: a short self-contained recap of the covariant-measurement properties invoked from quantum metrology would help readers who are not specialists in that sub-area.
[Concentration bounds section] The concentration bound exp(-Θ(N)t²) should explicitly state whether it applies in operator norm, Frobenius norm, or entrywise, and whether the hidden constants depend on the number of parameters in θ.
Notation for the CFIM F^U(θ) and the Haar measure μ_H should be introduced once with a clear definition before the variance calculation.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and accurate summary of our manuscript, the recognition of its potential utility for quantum natural gradient methods, and the recommendation for minor revision. The report correctly identifies the main technical contributions, including the exact variance scaling and non-asymptotic concentration bounds.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper reviews the known identity E_{U~μ_H}[F^U(θ)] = (1/2)Q(θ) for pure states by connecting it to covariant measurements in quantum metrology, then independently derives the exact variance O(N^{-1}), moment estimates, and non-asymptotic concentration bounds exp(-Θ(N)t²) from that established starting point. These steps do not reduce by construction to self-definitions, fitted parameters renamed as predictions, or load-bearing self-citations; the central claims build upon an externally attributed identity without circular reduction to the paper's own inputs or assumptions. The analysis remains explicitly restricted to pure states in C^N with no unsupported extrapolation.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central results rest on the assumption of pure states and the use of the Haar measure on the unitary group; no free parameters are introduced and no new entities are postulated.

axioms (2)

domain assumption The quantum state is pure and lies in C^N
The averaging identity E[F^U] = 1/2 Q is stated to hold specifically for pure states.
standard math Measurement bases are drawn from the Haar measure μ_H on the unitary group
The expectation and concentration statements are defined with respect to this invariant measure.

pith-pipeline@v0.9.0 · 5715 in / 1364 out tokens · 49832 ms · 2026-05-18T18:41:21.649283+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

averaging the classical Fisher information matrix over Haar-random measurement bases yields E_{U~μ_H}[F^U(θ)] = 1/2 Q(θ) for pure states in C^N

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

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

Quantum metrology of mixed states via purification
quant-ph 2026-05 unverdicted novelty 6.0

New purification-based reformulations of QCRB and HCRB connect mixed-state metrology bounds to those of purified states, enabling asymptotic attainment of HCRB or 2×QCRB via random channels and individual measurements.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages · cited by 1 Pith paper

[1]

2, 251–276

[Ama98] Shun-Ichi Amari,Natural gradient works efficiently in learning, Neural computation10(1998), no. 2, 251–276. [CAB+21] Marco Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C Benjamin, Suguru Endo, Keisuke Fujii, Jarrod R McClean, Kosuke Mitarai, Xiao Yuan, Lukasz Cincio, and Patrick Coles,Variational quantum algorithms, Na- ture Reviews Physics3(2021...

work page 1998
[2]

[KJN25] Ioannis Kolotouros, David Joseph, and Anand Kumar Narayanan,Accelerating quantum imaginary-time evo- lution with random measurements, Phys. Rev. A111(2025Jan), 012424. [KW24] Ioannis Kolotouros and Petros Wallden,Random Natural Gradient, Quantum8(October 2024),

work page 2024
[3]

26, 260501

[RBMV21] Aniket Rath, Cyril Branciard, Anna Minguzzi, and Benoît Vermersch,Quantum Fisher information from ran- domized measurements, Physical Review Letters127(2021), no. 26, 260501. 22 JIANFENG LU AND KECEN SHA [SCR07] Sandro Sorella, Michele Casula, and Dario Rocca,Weak binding between two aromatic rings: Feeling the van der Waals attraction by quantum...

work page 2021
[4]

[SIKC20] James Stokes, Josh Izaac, Nathan Killoran, and Giuseppe Carleo,Quantum natural gradient, Quantum4 (2020),

work page 2020
[5]

24, 241103

[Sor05] Sandro Sorella,Wave function optimization in the variational Monte Carlo method, Physical Review B—Condensed Matter and Materials Physics71(2005), no. 24, 241103. [Sor98] ,Green function Monte Carlo with stochastic reconfiguration, Physical review letters80(1998), no. 20,

work page 2005
[6]

[TAU16] Julien Toulouse, Roland Assaraf, and Cyrus J Umrigar,Introduction to the variational and diffusion Monte Carlo methods, Advances in quantum chemistry, 2016, pp. 285–314. [ZC25] Sisi Zhou and Senrui Chen,Randomized measurements for multi-parameter quantum metrology,

work page 2016
[7]

[ZH18] Huangjun Zhu and Masahito Hayashi,Universally Fisher-symmetric informationally complete measurements, Phys

arXiv:2502.03536. [ZH18] Huangjun Zhu and Masahito Hayashi,Universally Fisher-symmetric informationally complete measurements, Phys. Rev. Lett.120(2018Jan), 030404. DEPARTMENTS OFMATHEMATICS, PHYSICS,ANDCHEMISTRY, DUKEUNIVERSITY, DURHAM, NC 27708 USA Email address:jianfeng@math.duke.edu SCHOOL OFMATHEMATICALSCIENCES, PEKINGUNIVERSITY, BEIJING, 100871 CHIN...

work page arXiv

[1] [1]

2, 251–276

[Ama98] Shun-Ichi Amari,Natural gradient works efficiently in learning, Neural computation10(1998), no. 2, 251–276. [CAB+21] Marco Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C Benjamin, Suguru Endo, Keisuke Fujii, Jarrod R McClean, Kosuke Mitarai, Xiao Yuan, Lukasz Cincio, and Patrick Coles,Variational quantum algorithms, Na- ture Reviews Physics3(2021...

work page 1998

[2] [2]

[KJN25] Ioannis Kolotouros, David Joseph, and Anand Kumar Narayanan,Accelerating quantum imaginary-time evo- lution with random measurements, Phys. Rev. A111(2025Jan), 012424. [KW24] Ioannis Kolotouros and Petros Wallden,Random Natural Gradient, Quantum8(October 2024),

work page 2024

[3] [3]

26, 260501

[RBMV21] Aniket Rath, Cyril Branciard, Anna Minguzzi, and Benoît Vermersch,Quantum Fisher information from ran- domized measurements, Physical Review Letters127(2021), no. 26, 260501. 22 JIANFENG LU AND KECEN SHA [SCR07] Sandro Sorella, Michele Casula, and Dario Rocca,Weak binding between two aromatic rings: Feeling the van der Waals attraction by quantum...

work page 2021

[4] [4]

[SIKC20] James Stokes, Josh Izaac, Nathan Killoran, and Giuseppe Carleo,Quantum natural gradient, Quantum4 (2020),

work page 2020

[5] [5]

24, 241103

[Sor05] Sandro Sorella,Wave function optimization in the variational Monte Carlo method, Physical Review B—Condensed Matter and Materials Physics71(2005), no. 24, 241103. [Sor98] ,Green function Monte Carlo with stochastic reconfiguration, Physical review letters80(1998), no. 20,

work page 2005

[6] [6]

[TAU16] Julien Toulouse, Roland Assaraf, and Cyrus J Umrigar,Introduction to the variational and diffusion Monte Carlo methods, Advances in quantum chemistry, 2016, pp. 285–314. [ZC25] Sisi Zhou and Senrui Chen,Randomized measurements for multi-parameter quantum metrology,

work page 2016

[7] [7]

[ZH18] Huangjun Zhu and Masahito Hayashi,Universally Fisher-symmetric informationally complete measurements, Phys

arXiv:2502.03536. [ZH18] Huangjun Zhu and Masahito Hayashi,Universally Fisher-symmetric informationally complete measurements, Phys. Rev. Lett.120(2018Jan), 030404. DEPARTMENTS OFMATHEMATICS, PHYSICS,ANDCHEMISTRY, DUKEUNIVERSITY, DURHAM, NC 27708 USA Email address:jianfeng@math.duke.edu SCHOOL OFMATHEMATICALSCIENCES, PEKINGUNIVERSITY, BEIJING, 100871 CHIN...

work page arXiv