Pith. sign in

REVIEW 3 major objections 5 minor 41 references

Calibrated Pressure-Observable Born and Hessian Actions for Quantum-Assisted Waveform Inversion

T0 review · 3 major / 5 minor · reviewed 2026-07-30 · grok-4.5

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

Pith's one-line read Physical pressure in Schrödingerised acoustics has a two-term derivative; both terms are required for correct Born, adjoint, and Gauss–Newton FWI updates.

desk verdict Solid methods paper: the receiver-calibration term in D(cπ) is the real contribution, and the discrete checks back it hard; the Schrödingerised-to-physical consistency chain still leans on assumed auxiliary rates and a toy ideal-circuit demo. read the letter →

arxiv 2607.26880 v1 pith:FDPNTXRC submitted 2026-07-29 quant-ph cs.NAmath.NA

Guanyu Li , Jiwei Jia , Yu Wang , Yuping Duan This is my paper · ORCID
classification quant-phcs.NAmath.NA MSC 65M3265M1265M0681P6835L0565F22
keywords full-waveforminversionBornmodellingSchrödingerisationquantumalgorithmsinversescatteringHessianactionpressurecalibrationselected-outputmeasurement
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

Full-waveform inversion recovers wavespeed from measured pressure by repeatedly applying Born, adjoint, and Gauss–Newton actions. This paper shows how to build those actions from Schrödingerised wave propagation so that the physical pressure observable stays consistent. Energy variables give a Hamiltonian, but pressure equals wavespeed times one of those variables, so its derivative always has two pieces: a propagated wavefield sensitivity and a direct receiver-calibration term. Keeping both yields compatible discrete operators; dropping the calibration term leaves an order-one Born error and can send the regularized update in the wrong direction. The authors prove a conditional consistency estimate, cost the preparation and selected-output measurements, compile a nine-qubit Born circuit, and drive a four-parameter hybrid inversion in which all ten predeclared finite-shot runs cut the initial model error. The point is practical: Hamiltonian wave evolution alone does not define a local FWI step until the physical-pressure derivative and the measured update observables are specified.

What carries the argument

The calibrated pressure Born row: Duhamel differentiation of the Schrödingerised state plus the explicit receiver-row derivative ℓj(δc π0), assembled into matrix-free actions Jv, J⊤r, and J⊤Jv under the chosen data and model inner products.

What would settle it

On the paper’s smooth periodic tests, drop the receiver-calibration term and check whether the Born finite-difference error stays order one and the regularized Gauss–Newton step keeps relative error near 1.3 and correlation near 0.07 against the calibrated step; if those failures disappear without the term, the central claim is wrong.

Watch

Extended reading notes

Core claim

Differentiating physical pressure p = cπ at background c0 produces D(cπ)[c0](δc) = c0 δπ + δc π0. Retaining both the propagated term and the receiver-calibration term in the Born map, its adjoint, and the Gauss–Newton normal action gives pressure-consistent operators for Schrödingerised constant-density acoustics; omitting calibration differentiates a different observable and substantially changes the local update.

Load-bearing premise

The link from Schrödingerised propagation to a correct local update holds only under stated stability and auxiliary-recovery error budgets, and the hybrid demonstration uses ideal-circuit probabilities at four model parameters rather than noisy hardware at field scale.

Editorial extensions

If this is right

  • Quantum or hybrid FWI pipelines that simulate energy variables must still measure calibrated physical-pressure rows, not raw π amplitudes.
  • Selected sketches of receiver–time functionals can replace full-gather readout when the goal is a local gradient or Hessian-action direction.
  • Resource counts must bill receiver preparation, LCU normalization, and calibration-term estimation separately from Hamiltonian simulation.
  • Omitting the calibration block can make a line search reject the update even when the omitted-term solver is internally consistent.

Reading between the lines

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

  • The same product-rule gap will appear in any coefficient-dependent observable built on top of an energy-scaled or impedance-scaled Hamiltonian, not only acoustic pressure.
  • Once variable density, PML auxiliaries, or transducer models enter the generator, each adds its own explicit derivative rows that must be costed like the receiver-calibration term.
  • Selected-output measurement plus classical normal-system assembly suggests near-term hybrid loops may be limited more by how many Born columns and shots a local update needs than by full-gather tomography.
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 / 5 minor

Summary. The paper constructs a pressure-consistent Born, adjoint, and Gauss–Newton interface for constant-density acoustic FWI from Schrödingerised first-order dynamics. Because physical pressure is p=cπ rather than the energy variable π, the directional derivative retains both a propagated term c0 δπ and a direct receiver-calibration term δc π0; both are kept in the Duhamel Born map and its adjoint/normal actions. The authors give a conditional consistency budget (Props. 4.1–4.2), a component-wise resource model, multi-method discrete verification (FD, independent RK4 tangent/adjoint, autodiff JVP/VJP, explicit Jacobians), second-order periodic spatial refinement when calibration is retained versus an O(1) plateau when it is omitted, assembled-matrix statevector checks, and a nine-qubit compiled midpoint Born circuit whose ideal-circuit Bernoulli samples drive a four-parameter hybrid VQLS update in which all ten predeclared seeds reduce model error.

Significance. If the construction holds under the stated hypotheses, the paper supplies a concrete and previously under-specified operator-and-readout bridge between Schrödingerised wave propagation and the matrix-free actions a local FWI step actually needs. The strongest contribution is not a complexity claim but the calibrated pressure derivative itself: the ablation (order-one Born FD error; regularized GN step relative error 1.296 and correlation 0.068 when calibration is dropped) is falsifiable and load-bearing, and the discrete chain is checked by independent tangent/reverse codes, autodiff, and explicit Jacobians rather than by construction. The resource table separating access, preparation, LCU normalization, quadrature, and selected-output measurement is a useful accounting template. The hybrid demo and resource model are toy-scale and ideal-circuit, so significance for quantum advantage remains prospective; significance for correct pressure-observable differentiation in Schrödingerised FWI interfaces is real and well evidenced at the finite-dimensional level.

major comments (3)
  1. [§4.1–4.2, Prop. 4.2, Table 6.3] Props. 3.1 and 4.1–4.2 transfer auxiliary-recovery error into calibrated pressure Born rows under hypothesized bounds ε_R, CΔp², and Ce^{-αL_p}. Prop. 4.2 explicitly assumes those auxiliary rates for background and tangent states rather than deriving them for R_pa acting on Duhamel tangent fields. Table 6.3 shows clear second-order spatial behavior for the complete Born action, but the auxiliary row only reports a 3.3× reduction (N_p=3→11) as an “empirical trend,” with no fitted order and no separate ψ0 vs δψ breakdown. This is the soft link in the Schrödingerised-to-physical claim: internal discrete consistency (FD/autodiff/Jacobians) can hold while the continuous/semi-discrete Born interface retains an uncontrolled gap on tangent recovery. Please either (i) prove or numerically establish auxiliary rates for the tangent/Born map under the same norms as Prop. 4.2, or (ii) restate the mai
  2. [Abstract; §5.2; §6.2; §8] The compiled nine-qubit instance and hybrid inversion sample Bernoulli outcomes from ideal-circuit probabilities (unrestricted connectivity, classically precomputed Pauli coefficients, no device noise). Normal-system assembly, regularization, line search, and model refresh are classical; VQLS acts on a 4×4 normalized system. This is a legitimate measurement-chain prototype, but the abstract, title, and conclusion language (“quantum-assisted waveform inversion,” “connect Schrödingerised propagation to a local FWI update”) can be read as claiming a demonstrated quantum FWI pathway beyond what §5–6 and the Discussion actually show. Please calibrate claims to the evidence: ideal finite-shot selected-output interface at toy scale, with scalable oracles and hardware noise deferred. In particular, state explicitly that no asymptotic advantage or noise-robust update is claimed.
  3. [§2.1; §3.1 (Source convention); §6; Supplement S6 Table S10] The finite-dimensional diagnostics use a coefficient-independent first-order source-state convention (3.10), while the continuous fixed-physical-forcing map (2.3) contributes an extra (δc f_s, 0) term. Supplement Table S10 shows O(1) Born error when that source derivative is omitted from the fixed-physical-forcing map—an important audit—but the main text’s primary operator chain and the compiled circuit follow the fixed-state convention. For the central “physical-pressure derivative” claim, please state in the main narrative (not only the supplement) which convention the reported J, J^T, J^T J and the hybrid update implement, and that a fixed-physical-forcing realization must retain the source term analogously to receiver calibration. Otherwise readers may export the calibrated receiver construction while silently dropping a parallel O(1) source contribution.
minor comments (5)
  1. [Title page / header] The running header and title page show spaced/broken tokens (“CALIBRA TED”, “OBSER V ABLE”, “W A VEFORM”). Clean typography in the production version.
  2. [§6.1, Fig. 6.1] Figure 6.1 right panel and the omitted-calibration GN diagnostics are central; ensure the singular-spectrum comparison and the step-correlation/line-search rejection numbers (relative error 1.296, correlation 0.068) appear with explicit pointers in the main text near the figure, not only in the supplement.
  3. [§4.4, Table 4.1] Table 4.1 is helpful accounting but mixes asymptotic oracle language with “evidence in this paper” that is dense/statevector/compiled. A one-line caption note that the compiled instance precomputes Paulis (no sparse oracle) would avoid over-reading the table as an implemented block-encoding.
  4. [§3–4] Notation: L, L_ac_Sch, J, and bL appear for closely related maps; a small notation paragraph listing continuous vs discrete vs computed vs sketched objects would reduce load in §§3–4.
  5. [§1] Related-work pointers to classical adjoint-state and GN FWI are adequate; a brief sentence contrasting the present pressure-row derivative with frozen-observable or π-only readouts in prior Schrödingerisation PDE circuits would sharpen novelty without expanding scope.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Born/adjoint/GN actions are derived by product rule and Duhamel, then checked against independent discrete oracles rather than defined to match the target.

full rationale

The load-bearing analytic step is the product-rule identity D(cπ)[c0](δc)=c0 δπ+δc π0 (Prop. 2.1) plus Duhamel differentiation of the Schrödingerised propagator, which yields the calibrated Born row, its Euclidean adjoint, and the Gauss–Newton normal action. That chain does not define the operators in terms of the numerical targets they are later compared to. Consistency (Props. 4.1–4.2) is explicitly conditional on stated recovery/discretization budgets; those hypotheses are assumptions, not circular closures. Discrete verification uses finite differences, independently coded tangent/reverse RK4, autodiff JVP/VJP, and explicit Jacobians—external oracles relative to the claimed map. Ablations (omit receiver calibration; omit source derivative) produce O(1) Born errors and rejected GN steps, so the construction is falsifiable rather than true by definition. The four-parameter hybrid uses same-model synthetic data (standard inverse-crime toy setting) and ideal-circuit Bernoulli samples; that limits external claim strength but does not make the update equal its inputs by construction. No load-bearing self-citation uniqueness theorem or fitted-parameter-as-prediction pattern appears. Weakest links are assumed auxiliary-recovery rates, not circularity.

Assumptions & free parameters 4 free parameters · 6 assumptions · 2 invented entities

The work sits on standard acoustic FWI and Schrödingerisation plus explicit modeling choices (constant density, fixed interfaces, energy variables, ideal circuits). Free parameters are algorithmic knobs for the toy hybrid demo, not physical constants fitted to nature. Invented entities are definitional interfaces (calibrated pressure observable, selected-output measurement), not new physics particles.

free parameters (4)
  • GN regularization λ_k = 1e-4 * max(tr(J^T J)/4, 1e-14) = 1e-4 relative to average eigenvalue proxy
    Hand-chosen regularization schedule for the four-parameter local system; affects accepted steps and reported correlations.
  • VQLS/SPSA hyperparameters (a=0.5, c=0.12, 1000 iters, 5000 shots/term) = a_SPSA=0.5, c_SPSA=0.12, 1000 iterations
    Optimizer and shot budget chosen for the two-qubit demo; finite-shot residual ~0.11 vs machine-precision classical solve.
  • Product-formula repetitions r and Duhamel nodes Q = compiled prototype Q=1, r=4
    Discretization knobs (e.g. r=4, Q=1 midpoint in compiled instance) controlling simulation/quadrature error in the resource and circuit sections.
  • Line-search acceptance rule (3 fresh forward estimates vs incumbent) = N_rep=3, N_α=5
    Conservative finite-shot acceptance rule that produced two late α=0 decisions; procedural choice, not derived.
assumptions (6)
  • domain assumption Constant-density acoustic wave model with pressure identified as ∂t u (up to scaling) and energy variables π=c^{-1}∂t u, q=∇u.
    Section 2; excludes variable density and changes Jacobian factors under other parameterizations.
  • domain assumption Schrödingerisation embedding yields a Hermitian Hamiltonian on extended space with recovery map R_pa whose background and derivative errors are bounded by ε_R (Prop. 3.1).
    Section 3; relies on Jin et al. Schrödingerisation framework plus stated recovery hypotheses.
  • domain assumption Fixed material interfaces / pixel models; no shape derivatives for moving discontinuities.
    Section 2 fixed-interface derivative convention; high-contrast geology treated as grid values.
  • standard math Uniform stability of continuous and semidiscrete background/tangent families and second-order consistency of centered periodic differences on smooth C^3 coefficients (Prop. 4.2).
    Standard FD/PDE stability assumptions used to get O(h^2) spatial rates.
  • ad hoc to paper Hybrid demo samples Bernoulli outcomes from ideal compiled-circuit probabilities (no device noise, unrestricted connectivity).
    Sections 5–6 and Discussion; explicitly deferred hardware-mapped noise analysis.
  • domain assumption Selected-output sketches or few pressure/Born scalars suffice to preserve local update directions for the tested maps.
    Section 4.4 and measurement experiments; empirical on small systems, not a general theorem.
invented entities (2)
  • Calibrated pressure-observable Born map L_ac_Sch with explicit receiver-calibration term DΓ[c0](δc) independent evidence
    purpose: Make Schrödingerised evolution differentiate the physical pressure p=cπ rather than frozen c0 π.
    Definitional operator interface (Prop. 2.1, Eqs. 3.13–3.14); standard product rule, not new physics.
  • Selected-output measurement target for local FWI (sketch S of receiver–time functionals) independent evidence
    purpose: Avoid full-gather readout cost while preserving gradient/Hessian-action directions.
    Measurement-model choice in §3.4–4.4; empirically tested on small Jacobians.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Calibrated Pressure-Observable Born and Hessian Actions for Quantum-Assisted Waveform Inversion." pith.science (2026). https://pith.science/paper/FDPNTXRC

@misc{pith2026260726880,
  author       = {Pith},
  title        = {Pith review of: Calibrated Pressure-Observable Born and Hessian Actions for Quantum-Assisted Waveform Inversion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FDPNTXRC}},
  note         = {Machine review of arXiv:2607.26880}
}
abstract

We construct a pressure-consistent operator-and-readout interface for Born, adjoint, and Gauss--Newton actions in constant-density acoustic full-waveform inversion (FWI) using Schr\"odingerised propagation. The energy variables $\pi=c^{-1}\partial_t u$ and $q=\nabla u$ yield an auxiliary-space Hamiltonian, while physical pressure $p=c\pi$ depends explicitly on wavespeed. Its derivative $D(c\pi)[c_0](\delta c)=c_0\delta\pi+\delta c\,\pi_0$ combines propagated wavefield sensitivity with a direct receiver-calibration term. Duhamel and receiver-row differentiation retain both contributions in the Born map, its adjoint, and the Gauss--Newton normal action. We prove a conditional consistency estimate with a periodic second-order finite-difference specialization and give a resource model for state preparation, normalization, quadrature, and selected-output measurement. A compiled nine-qubit instance realizes structured preparation, product-formula propagation, a derivative-LCU block, and calibrated pressure-overlap measurements. Bernoulli samples from ideal-circuit probabilities drive a four-parameter hybrid inversion. A two-qubit VQLS circuit represents the normalized update direction, while normal-system assembly, line search, and model refresh remain classical. Finite differences, tangent and reverse-adjoint recurrences, autodiff JVP/VJP evaluations, and explicit Jacobians verify the discrete Born, adjoint, and normal actions. Smooth periodic refinement confirms second-order convergence, whereas omitting receiver calibration leaves an order-one Born error and substantially changes the regularized Gauss--Newton direction. All ten predeclared finite-shot runs reduce the initial model error. These results specify the physical-pressure derivative and selected-output measurements needed to connect Schr\"odingerised propagation to a local FWI update.

Figures

Figures reproduced from arXiv: 2607.26880 by the authors.

Figure 6
Figure 6. [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 6
Figure 6. [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references

  1. [1]

    Conference on Inverse Scattering: Theory and Application , editor =

    Lailly, Patrick , title =. Conference on Inverse Scattering: Theory and Application , editor =. 1983 , pages =

    1983

  2. [2]

    Geophysics , year =

    Tarantola, Albert , title =. Geophysics , year =

  3. [3]

    and Zaleski, St

    Bunks, Carey and Saleck, Fatimetou M. and Zaleski, St. Multiscale seismic waveform inversion , journal =. 1995 , volume =

    1995

  4. [4]

    Gerhard and Shin, Changsoo and Hicks, Graham J

    Pratt, R. Gerhard and Shin, Changsoo and Hicks, Graham J. , title =. Geophys. J. Int. , year =

  5. [5]

    Full Waveform Inversion and the Truncated

    M. Full Waveform Inversion and the Truncated. SIAM J. Sci. Comput. , year =

  6. [6]

    A review of the adjoint-state method for computing the gradient of a functional with geophysical applications , journal =

    Plessix, Ren. A review of the adjoint-state method for computing the gradient of a functional with geophysical applications , journal =. 2006 , volume =

    2006

  7. [7]

    An overview of full-waveform inversion in exploration geophysics , journal =

    Virieux, Jean and Operto, St. An overview of full-waveform inversion in exploration geophysics , journal =. 2009 , volume =

    2009

  8. [8]

    and Bourget, M

    Brougois, A. and Bourget, M. and Lailly, P. and Poulet, M. and Ricarte, P. and Versteeg, R. , title =. EAEG Workshop---Practical Aspects of Seismic Data Inversion , publisher =. 1990 , doi =

    1990

Show all 41 references
  1. [9]

    Alfarhan, Mustafa and Ravasi, Matteo and Chen, Fuqiang and Alkhalifah, Tariq , title =. Geophys. J. Int. , year =

  2. [10]

    , title =

    Wang, Kun and Matthews, Thomas and Anis, Fatima and Li, Cuiping and Duric, Neb and Anastasio, Mark A. , title =. IEEE Trans. Ultrason. Ferroelectr. Freq. Control , year =

  3. [11]

    High Resolution

    Lucka, Felix and P. High Resolution. Inverse Problems , year =

  4. [12]

    and Saunders, Michael A

    Paige, Christopher C. and Saunders, Michael A. , title =. ACM Trans. Math. Software , year =

  5. [13]

    Numerical Methods for Least Squares Problems , publisher =

    Bj. Numerical Methods for Least Squares Problems , publisher =. 1996 , doi =

    1996

  6. [14]

    and Nocedal, Jorge , title =

    Liu, Dong C. and Nocedal, Jorge , title =. Math. Program. , year =

  7. [15]

    and Sheikh, Hamid R

    Wang, Zhou and Bovik, Alan C. and Sheikh, Hamid R. and Simoncelli, Eero P. , title =. IEEE Trans. Image Process. , year =

  8. [16]

    Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2015 , series =

    Ronneberger, Olaf and Fischer, Philipp and Brox, Thomas , title =. Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2015 , series =. 2015 , doi =

    2015

  9. [17]

    and Hassidim, Avinatan and Lloyd, Seth , title =

    Harrow, Aram W. and Hassidim, Avinatan and Lloyd, Seth , title =. Phys. Rev. Lett. , year =

  10. [18]

    and Childs, Andrew M

    Berry, Dominic W. and Childs, Andrew M. and Kothari, Robin , title =. Proceedings of the 56th Annual. 2015 , pages =

    2015

  11. [19]

    and Kothari, Robin and Somma, Rolando D

    Childs, Andrew M. and Kothari, Robin and Somma, Rolando D. , title =. SIAM J. Comput. , year =

  12. [20]

    and Childs, Andrew M

    Berry, Dominic W. and Childs, Andrew M. and Ostrander, Aaron and Wang, Guoming , title =. Comm. Math. Phys. , year =

  13. [21]

    and Liu, Jin-Peng and Ostrander, Aaron , title =

    Childs, Andrew M. and Liu, Jin-Peng and Ostrander, Aaron , title =. Quantum , year =

  14. [22]

    Linden, Noah and Montanaro, Ashley and Shao, Changpeng , title =. Comm. Math. Phys. , year =

  15. [23]

    Explicit Quantum Circuits for Block Encodings of Certain Sparse Matrices , journal =

    Camps, Daan and Lin, Lin and. Explicit Quantum Circuits for Block Encodings of Certain Sparse Matrices , journal =. 2024 , volume =

    2024

  16. [24]

    Danz, Sven and Stollenwerk, Tobias and Ciani, Alessandro , title =. SIAM J. Sci. Comput. , year =

  17. [25]

    Quantum , year =

    An, Dong and Fang, Di and Lin, Lin , title =. Quantum , year =

  18. [26]

    , title =

    Low, Guang Hao and Chuang, Isaac L. , title =. Quantum , year =

  19. [27]

    and Subasi, Yigit and Cincio, Lukasz and Coles, Patrick J

    Bravo-Prieto, Carlos and LaRose, Ryan and Cerezo, M. and Subasi, Yigit and Cincio, Lukasz and Coles, Patrick J. , title =. Quantum , year =

  20. [28]

    and Chuang, Isaac L

    Nielsen, Michael A. and Chuang, Isaac L. , title =. 2010 , doi =

    2010

  21. [29]

    Quantum amplitude amplification and estimation , booktitle =

    Brassard, Gilles and H. Quantum amplitude amplification and estimation , booktitle =. 2002 , pages =

    2002

  22. [30]

    Costa, Pedro C. S. and Jordan, Stephen P. and Ostrander, Aaron , title =. Phys. Rev. A , year =

  23. [31]

    Quantum computing for fluids: where do we stand? , journal =

    Succi, Sauro and Itani, Wael and Sreenivasan, Katepalli and Steijl, Ren. Quantum computing for fluids: where do we stand? , journal =. 2023 , volume =

    2023

  24. [32]

    Jin, Shi and Liu, Nana and Yu, Yue , title =. Phys. Rev. A , year =

  25. [33]

    Jin, Shi and Liu, Nana and Yu, Yue , title =. Phys. Rev. Lett. , year =

  26. [34]

    2025 , note =

    Jin, Shi and Zhang, Chundan , title =. 2025 , note =

    2025

  27. [35]

    Jin, Shi and Li, Xiantao and Liu, Nana and Yu, Yue , title =. J. Comput. Phys. , year =

  28. [36]

    Quantum , year =

    Hu, Junpeng and Jin, Shi and Liu, Nana and Zhang, Lei , title =. Quantum , year =

  29. [37]

    2026 , note =

    Nguyen, Hoang Anh and Vashisth, Divakar and Tura, Ali , title =. 2026 , note =

    2026

  30. [38]

    2025 , note =

    Sato, Yuki and Kato, Jumpei and Yano, Hiroshi and Ito, Kosuke and Yamamoto, Naoki , title =. 2025 , note =

    2025

  31. [39]

    and Lishman, Jake and Gacon, Julien and Martiel, Simon and Nation, Paul D

    Javadi-Abhari, Ali and Treinish, Matthew and Krsulich, Kevin and Wood, Christopher J. and Lishman, Jake and Gacon, Julien and Martiel, Simon and Nation, Paul D. and Bishop, Lev S. and Cross, Andrew W. and Johnson, Blake R. and Gambetta, Jay M. , title =. 2024 , note =

    2024

  32. [40]

    and Stiefel, Eduard , title =

    Hestenes, Magnus R. and Stiefel, Eduard , title =. J. Res. Natl. Bur. Stand. , year =

  33. [41]

    , title =

    Spall, James C. , title =. IEEE Trans. Automat. Control , volume =. 1992 , doi =

    1992

Pith tools

Reviewed July 30, 2026 · model on record in the stance chip above.