Pith. sign in

REVIEW 3 major objections 3 minor 119 references

Stochastic Pauli-path simulator for large-scale quantum optimization

T0 review · 3 major / 3 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read By sampling instead of truncating Pauli propagation paths, SPPS obtains unbiased gradients and provably convergent optimization trajectories.

desk verdict Strong, novel PBS-gradient framework with a clean value-gradient separation theorem, but the deployed adaptive stopping rule escapes the unbiasedness proof, so the guarantees apply to a different algorithm than the one that produced the experiments. read the letter →

arxiv 2607.17804 v1 pith:3FVHKU56 submitted 2026-07-20 quant-ph

Kaining Zhang , Xinbiao Wang , Kunsheng Li , Qixin Zhang , Yuxuan Du , Min-Hsiu Hsieh , Dacheng Tao This is my paper · ORCID
classification quant-ph MSC 81P6868Q12 PACS 03.67.Lx03.67.Ac
keywords stochasticPauli-pathsimulationunbiasedgradientestimationvariationalquantumeigensolverneuralnetworksimportancereweightingpathautomaticdifferentiationPaulipropagationclassicalofoptimization
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

This paper argues that the standard way to make Pauli-based quantum simulators practical—truncating the explosion of propagation paths—is fundamentally incompatible with gradient-based optimization: a path set can approximate the objective value to any accuracy while its gradient is still order-one wrong (Theorem 1), and the error accumulates across descent steps into a macroscopic gap (Corollary 1). To fix this, the paper introduces the stochastic Pauli-path simulator (SPPS), which samples paths from the full path space, reweights each contribution by its sampling probability, and differentiates each path's trigonometric weight to form stochastic gradients. The central theoretical result is that these gradients are unbiased, with sample complexity governed by an effective branching factor $\kappa(\theta)=\prod_{j=1}^L(1+|\sin 2\theta_j|)$, and that SPPS-driven gradient descent reaches an $\epsilon$-stationary point in $T=O(L^3\|O\|_2^4/\epsilon^2)$ iterations. Numerically, SPPS tracks exact gradient descent on variational quantum eigensolver pre-training up to 100 qubits and quantum neural network pre-training up to 40 qubits, while running in minutes and beating truncation-based alternatives in both accuracy and runtime. A sympathetic reader should take the paper as establishing that stochastic path sampling, not deterministic truncation, is the right way to bring Pauli-based simulation into optimization tasks.

What carries the argument

The load-bearing object is the Pauli propagation path $\omega\in\{0,\pm1\}^L$: each entry records whether a rotation gate branches into its cosine or sine component, and the observable expectation expands as a sum over paths of $\Psi_\omega(\theta)\mathrm{Tr}[P_\omega(O)\rho]$. SPPS's sampling distribution assigns the two branches probabilities $q_j=(|\cos\theta_j|+a)/(|\cos\theta_j|+|\sin\theta_j|+2a)$ for a smoothing parameter $a>0$, so derivative-sensitive branches with tiny coefficients are still visited. Each sampled path is reweighted by $1/\Pr(\omega)$, and path automatic differentiation multiplies the reweighted contribution by score factors $-\tan\theta_j$ for $\omega_j=1$ and $\cot\theta_j$ for $\omega_j=-1$, yielding all gradient entries from the same samples. The quantity that controls cost is the effective branching factor $\kappa(\theta)=\prod_{j=1}^L(1+|\sin 2\theta_j|)$, which appears in the variance, sample complexity, and convergence bounds.

What would settle it

Fix a circuit and parameter point with known exact gradient, run the adaptive-stopping SPPS many times from independent seeds, and compare the distribution of the stopped averaged estimator to the exact gradient: a systematic offset larger than the Monte Carlo standard error, or a correlation between stopping time and estimator magnitude, would show the practical rule breaks the unbiasedness the theorems assume.

Watch

Extended reading notes

Core claim

The paper's central claim is that the gradient bias of truncation-based Pauli simulators is not incidental but structural. Theorem 1 exhibits, for any target accuracy, a truncated path set that reproduces the function value to within $\epsilon$ while missing a path whose derivative is order one, so the gradient error stays above $1-\epsilon$; Corollary 1 shows this bias pushes gradient descent to a wrong limit point separated by a macroscopic gap. Against that, SPPS samples legal Pauli propagation paths with a smoothed distribution, divides each sampled contribution by its probability, and computes all gradient components at once through path automatic differentiation, giving an unbiased stochastic gradient. Theorem 2 bounds the sample budget to $\tilde{O}((1+2a)^L L \kappa(\theta)/(a\epsilon^2))$ for an $\epsilon$-accurate gradient with high probability, and Corollary 2 converts unbiasedness plus a variance bound into a polynomial convergence guarantee for SPPS-driven descent. The experiments are the concrete payoff: faithful tracking of exact optimization trajectories on VQE and QNN benchmarks, with 100-qubit VQE pre-training completed in about a minute.

Load-bearing premise

The load-bearing premise is that the implemented adaptive sample-budget rule, which stops when two independent macro-replicates agree, is free of systematic error; the paper explicitly calls that stopping proxy practical rather than a theorem-level confidence bound, so if stopping bias sneaks into the average, the unbiased-gradient and convergence guarantees no longer cover the deployed code.

Editorial extensions

If this is right

  • Truncation-based Pauli simulators should not be trusted for gradient-driven optimization even when their energy estimates look accurate; a small value error can hide an order-one gradient error.
  • When the effective branching factor $\kappa(\theta)$ stays moderate along the optimization trajectory, SPPS provides polynomial sample budgets, so optimization dynamics join the class of classically simulable quantum processes.
  • SPPS-driven gradient descent converges to an approximate stationary point in $O(L^3\|O\|_2^4/\epsilon^2)$ iterations, giving the first convergence guarantee for a non-exact classical simulator of quantum optimization.
  • The same sampled paths yield both expectation values and all gradient components, so the cost of a gradient step is comparable to one forward simulation rather than a parameter-shift factor of $L$.
  • In practice the method pre-trains VQE on 100-qubit transverse-field Ising models in about one minute and a 40-qubit quantum neural network in under ten minutes while following exact optimization dynamics.

Reading between the lines

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

  • Editorial inference: the same unbiased-sampling strategy should extend to other propagation bases, such as Majorana fermions or Clifford perturbation theory, because the argument only needs a finite path expansion and a differentiable trigonometric weight per path.
  • Editorial inference: SPPS's unbiased gradient noise gives a direct classical probe of landscape flatness; the variance bound proportional to $\kappa(\theta)$ could be used to detect barren-plateau regimes before expensive quantum training.
  • Editorial inference: a clean testable consequence is that fixed-budget SPPS and adaptive-stopping SPPS should agree statistically; if they do not, the practical implementation carries a selection bias that the theorems do not cover.
  • Editorial inference: the value-gradient separation theorem suggests a new reporting standard for classical simulators: validate gradients, not just energies, on optimization benchmarks.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 manuscript introduces SPPS, a stochastic Pauli-path simulator that samples Heisenberg-propagation paths and uses importance reweighting plus 'path automatic differentiation' to estimate gradients of expectation values. It proves a value-gradient separation result for truncation-based PBS, establishes unbiasedness and a high-probability sample-complexity bound for a fixed-budget version of SPPS, and derives an SGD convergence corollary. The experiments cover VQE pre-training up to 100 qubits, QNN pre-training up to 40 qubits, and small state-preparation circuits, reporting trajectory tracking and runtime advantages over CT-, FT-, and WT-PBS baselines.

Significance. If the guarantees hold for the deployed algorithm, this is a meaningful step: Pauli-based simulation would move from forward estimation to optimization, with clear practical value for low-magic variational circuits. The paper has real strengths: the adversarial value-gradient separation in Theorem 1 is explicit and clean; the fixed-budget unbiasedness proof in App. E is self-contained; and the experiments are broad, including a 100-qubit VQE benchmark and a reproducibility link. The main caveat is that the estimator actually used in the experiments includes an adaptive stopping rule that lies outside the theorem-level guarantees, and the main-text sample-complexity statement is internally inconsistent with the appendix. These issues need to be resolved before the headline claims are fully supported.

major comments (3)
  1. [Section IV.B, Theorem 2; Appendix E, Eq. (E27)] The main-text Theorem 2 states a sample complexity of O-tilde((1+2a)^L kappa(theta)^L/(a epsilon^2)), but the formal Theorem 5 in Appendix E, Eq. (E27), gives B = O(L kappa(theta)(1+2a)^L log(L/delta)/(a epsilon^2)). Since kappa(theta) is already a product over the L layers, the extra exponent L in the main text is material and makes the two statements disagree. This internal inconsistency must be corrected, and the corrected version should be used in the discussion of when the sampling budget is polynomial.
  2. [Appendix D.5, Eqs. (D27)-(D31); Section IV.B, Theorem 2] The adaptive stopping rule described in App. D.5 is not covered by the theoretical guarantees. Theorem 2 and Appendix E fix the sample budget B before sampling, whereas the deployed estimator doubles the budget until the data-dependent proxy Delta-hat falls below delta and then returns a conditional average of the two macro-replicates. The identity E[Delta-hat^2] = E[||g~-g||^2] in Eqs. (D30)-(D31) is a fixed-budget statement; it does not imply E[g~ | Delta-hat <= delta] = g. For the skewed, heavy-tailed importance-reweighted path estimators used here, the selection event can bias the conditional mean. Because all reported experiments use the adaptive rule, Corollary 2's convergence guarantee does not currently apply to the evaluated simulator. The authors should either prove a bias or concentration statement for the stopped estimator or rerun the headline benchmarks with fixed budgets and report whether the results change.
  3. [Appendix D.3, Eq. (D17); Section IV.A, Eq. (6)] The derivation of path automatic differentiation in App. D.3 differentiates the reweighted contribution h~_omega(theta) = Psi_omega(theta)/Pr(omega) with respect to theta_j while treating Pr(omega) as constant. However, Pr(omega) depends on theta through q_j in Eq. (D5), so Eq. (D17) is not the derivative of h~_omega. The formal estimator in Appendix E, Eq. (E1), is defined directly as nabla Psi_omega/Pr(omega), which is unbiased, and the score identity in Eq. (6) follows from that definition. The appendix should be corrected so the derivation and the formal estimator agree; otherwise a reader following Eq. (D17) cannot reproduce the main-text formula.
minor comments (3)
  1. [Section IV.B, Theorem 2] The main text should state the restriction epsilon <= sqrt(L kappa(theta)) that appears in the formal Appendix E statement of the high-probability bound, since the stated formula only guarantees the epsilon error in that regime.
  2. [Appendix G.3] The state-preparation benchmark uses a truncated Pauli expansion of the projector observable (keeping the largest 100 non-identity coefficients). This is an additional approximation beyond the full-O expansion in Theorem 2; the text should clarify that this benchmark demonstrates the estimator under observable truncation rather than the untruncated theorem.
  3. [Throughout] The text contains numerous spacing and formatting inconsistencies involving 'Tb-PBS', 'CT-PBS', and related acronyms, and some figure labels (especially Fig. 1) are very hard to read in the provided version. A careful copyedit would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the SPPS derivation is self-contained.

full rationale

The central claims of the paper are derived from first principles without circular steps. The unbiasedness of the SPPS gradient estimator follows directly from importance reweighting of the exact Pauli expansion: Eq. (D6) defines ρ-weighted contributions and Eq. (D7) computes the expectation over sampled paths as the exact value f(θ); Eq. (D22) shows the same for gradients via the log-derivative trick. These are algebraic identities, not fitted relationships. The sample-complexity guarantee in Theorem 2 is proven in Appendix E by bounding the single-sample variance (Lemma 1) and applying Bernstein's inequality; the expression κ(θ) is a property of the circuit and parameters, not a fitted constant. Corollary 2 is a standard variance-based SGD convergence proof (Lemma 4) applied to the derived variance bound, with no hidden circular input. Theorem 1 and Corollary 1 are constructive counterexamples, not assumptions. Self-citations appear only for background, baseline implementations, or auxiliary techniques (e.g., PauliPropagation.jl), and none is load-bearing for the paper's core theorems. The adaptive A/B stopping rule in Appendix D.5 is explicitly labeled 'not used as a theorem-level confidence bound'; this is an acknowledged gap between the idealized fixed-budget theorem and the deployed heuristic, but it is not a circular step because the paper does not claim theorem-level unbiasedness for the adaptive estimator. The smoothing parameter a and threshold δ are algorithm hyperparameters, not fitted to reproduce the benchmark outcomes; small-system VQE results are validated against exact GD from PennyLane, providing independent external grounding. Overall, the derivation chain does not reduce any prediction to its inputs by construction.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical entities are postulated. SPPS is algorithmic: path sampling, importance reweighting, PAD, and the adaptive proxy are procedures, not new particles, forces, or conserved quantities.

free parameters (2)
  • smoothing parameter a = a_init=0.25/L (VQE), a0=0.01 (QNN), b_a=0.5
    Controls branch sampling probabilities q_j=(|cos theta|+a)/(|cos theta|+|sin theta|+2a). Required positive for unbiasedness; variance bound scales as 1/a. Chosen per task.
  • gradient-error proxy threshold delta = delta in {2^-1, 2^-2, 2^-3}
    Stops adaptive sample budget per Pauli term; smaller delta yields more samples and smaller empirical gradient error. Not fixed by theory.
assumptions (4)
  • standard math Heisenberg propagation identity (App. A3): conjugating a Pauli P by a Pauli rotation gives one branch if it commutes and a cosine/sine split if it anti-commutes; Clifford layers map Pauli to Pauli.
    Used to define the exact path expansion in Eq. (A8)-(A12), the foundation for both Tb-PBS and SPPS.
  • domain assumption The circuit U(theta) consists only of Pauli rotations and Clifford gates (Eq. A4).
    This is the scope of the algorithm and all experiments; non-Clifford gates are not treated.
  • domain assumption Input states used in the theorems and benchmarks are computational-basis product states, so Tr[P_omega rho] is zero or +/-1.
    Theorems 1 and Corollary 1 assume rho=(|0><0|)^(⊗n); the VQE/QNN tasks use product or computational-basis inputs, making trace evaluation cheap.
  • ad hoc to paper The adaptive stopping proxy delta-hat based on two macro-replicates estimates the stochastic scale of the averaged gradient.
    App. D5 uses this proxy to allocate samples, and explicitly disclaims a theorem-level confidence bound; the deployed unbiasedness therefore rests on an unproven assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Stochastic Pauli-path simulator for large-scale quantum optimization." pith.science (2026). https://pith.science/paper/3FVHKU56

@misc{pith2026260717804,
  author       = {Pith},
  title        = {Pith review of: Stochastic Pauli-path simulator for large-scale quantum optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3FVHKU56}},
  note         = {Machine review of arXiv:2607.17804}
}
read the original abstract

Pauli-based simulators offer a promising route to large-scale classical simulation of quantum circuits in the low-magic regime. Yet their applicability remains largely limited to forward simulation, making them inadequate for optimization-driven quantum tasks such as variational state preparation and parameter initialization. Existing approaches either lack native support for gradient-based optimization or suffer from severe gradient bias. Here we propose the stochastic Pauli-path simulator (SPPS), a computational framework for large-scale quantum optimization that enables unbiased stochastic gradient estimation via Pauli-path sampling across optimization iterations. Our theoretical analysis shows that the proposed simulator yields unbiased gradient estimates and admits provable convergence guarantees. We systematically evaluate our proposal, including quantum eigensolver benchmarks with up to 100 qubits and quantum neural network benchmarks with up to 40 qubits. Across these tasks, SPPS faithfully tracks optimization dynamics, converges within minutes, and broadens the role of Pauli-based simulation from forward estimation to large-scale quantum optimization.

Figures

Figures reproduced from arXiv: 2607.17804 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p041_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p042_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p042_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8 [PITH_FULL_IMAGE:figures/full_fig_p043_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

119 extracted references · 53 canonical work pages

  1. [1]

    Implementation details ofSPPS 25

    Explicit case with standard truncation rules 20 C.Tb-PBSleads to sub-optimal convergence (Proof of Corollary 1) 22 D. Implementation details ofSPPS 25

    work page

  2. [2]

    Basics of quantum computing 15

    work page

  3. [3]

    Pauli path propagation in the Heisenberg picture 16

    work page

  4. [4]

    Related work 17 B.Tb-PBSleads to biased gradient (Proof of Theorem 1) 18

    work page

  5. [5]

    Adaptive gradient-error proxy 28

    work page

  6. [6]

    Implementation summary 29 E.SPPSgenerates unbiased gradients (Proof of Theorem 2) 29

    work page

  7. [24]

    Rudolph, Ross Duncan, Ivan Rungger, and Cristina Cîrstoiu

    Enrico Fontana, Manuel S. Rudolph, Ross Duncan, Ivan Rungger, and Cristina Cîrstoiu. Classical simulations of noisy variational quantum circuits.npj Quantum Information, 11(1):84, May 2025. 11

    2025

  8. [25]

    Ignacio Cirac, and Rahul Trivedi

    Guillermo González-García, J. Ignacio Cirac, and Rahul Trivedi. Pauli path simulations of noisy quantum circuits beyond average case.Quantum, 9:1730, May 2025

    2025

Show all 119 references
  1. [26]

    Efficient simulation of parametrized quantum circuits under nonunital noise through pauli backpropagation.Phys

    Victor Martinez, Armando Angrisani, Ekaterina Pankovets, Omar Fawzi, and Daniel Stilck França. Efficient simulation of parametrized quantum circuits under nonunital noise through pauli backpropagation.Phys. Rev. Lett., 134:250602, Jun 2025

    2025

  2. [27]

    Simulating noisy variational quantum algorithms: A polynomial approach.Phys

    Yuguo Shao, Fuchuan Wei, Song Cheng, and Zhengwei Liu. Simulating noisy variational quantum algorithms: A polynomial approach.Phys. Rev. Lett., 133:120603, Sep 2024

    2024

  3. [28]

    Rudolph, M

    Armando Angrisani, Alexander Schmidhuber, Manuel S. Rudolph, M. Cerezo, Zoë Holmes, and Hsin-Yuan Huang. Classically estimating observables of noiseless quantum circuits.Phys. Rev. Lett., 135:170602, Oct 2025

    2025

  4. [29]

    Cheng-Ju Lin, Hrant Gharibyan, and Vincent P. Su. Utility-Scale Quantum State Preparation: Classical Training using Pauli Path Simulation.Quantum, 10:2014, March 2026

    2014

  5. [30]

    Symbolic pauli propagation for gradient- enabled pre-training of quantum circuits.arXiv preprint arXiv:2512.16674, 2025

    Saverio Monaco, Jamal Slim, Florian Rehm, Dirk Krücker, and Kerstin Borras. Symbolic pauli propagation for gradient- enabled pre-training of quantum circuits.arXiv preprint arXiv:2512.16674, 2025

    2025 arXiv

  6. [31]

    Pauli propagation: A computational framework for simulating quantum systems.arXiv preprint arXiv:2505.21606, 2025

    Manuel S Rudolph, Tyson Jones, Yanting Teng, Armando Angrisani, and Zoë Holmes. Pauli propagation: A computational framework for simulating quantum systems.arXiv preprint arXiv:2505.21606, 2025

    2025 arXiv

  7. [32]

    Rudolph, Zoë Holmes, Lukasz Cincio, and M

    Pablo Bermejo, Paolo Braccia, Manuel S. Rudolph, Zoë Holmes, Lukasz Cincio, and M. Cerezo. Quantum convolutional neural networks are effectively classically simulable.PRX Quantum, 7:020304, Apr 2026

    2026

  8. [33]

    Clayton Peacock, and Dries Sels

    Nicolas Loizeau, J. Clayton Peacock, and Dries Sels. Quantum many-body simulations with PauliStrings.jl.SciPost Phys. Codebases, page 54, 2025

    2025

  9. [34]

    Circuit compression for 2d quantum dynamics.arXiv preprint arXiv:2507.01883, 2025

    Matteo D’Anna, Yuxuan Zhang, Roeland Wiersema, Manuel S Rudolph, and Juan Carrasquilla. Circuit compression for 2d quantum dynamics.arXiv preprint arXiv:2507.01883, 2025

    2025

  10. [35]

    Simulation of fermionic circuits using majorana propagation.arXiv preprint arXiv:2503.18939, 2025

    Aaron Miller, Joachim Favre, Zoë Holmes, Özlem Salehi, Rahul Chakraborty, Anton Nykänen, Zoltán Zimborás, Adam Glos, and Guillermo García-Pérez. Simulation of fermionic circuits using majorana propagation.arXiv preprint arXiv:2503.18939, 2025

    2025

  11. [36]

    Interplay of resources for universal continuous-variable quantum computing.arXiv preprint arXiv:2502.07670, 2025

    Varun Upreti and Ulysse Chabaud. Interplay of resources for universal continuous-variable quantum computing.arXiv preprint arXiv:2502.07670, 2025

    2025 arXiv

  12. [37]

    When quantum resources backfire: Non-gaussianity and symplectic coherence in noisy bosonic circuits.arXiv preprint arXiv:2510.07264, 2025

    Varun Upreti, Ulysse Chabaud, Zoë Holmes, and Armando Angrisani. When quantum resources backfire: Non-gaussianity and symplectic coherence in noisy bosonic circuits.arXiv preprint arXiv:2510.07264, 2025

    2025

  13. [38]

    Fast convergence of majorana propagation for weakly interacting fermions

    Giorgio Facelli, Hamza Fawzi, and Omar Fawzi. Fast convergence of majorana propagation for weakly interacting fermions. arXiv preprint arXiv:2601.05226, 2026

    2026

  14. [39]

    Fast and converged classical simulations of evidence for the utility of quantum computing before fault tolerance.Science Advances, 10(3):eadk4321, 2024

    Tomislav Begušić, Johnnie Gray, and Garnet Kin-Lic Chan. Fast and converged classical simulations of evidence for the utility of quantum computing before fault tolerance.Science Advances, 10(3):eadk4321, 2024

    2024

  15. [40]

    Pauli propagation for imaginary time evolution.arXiv preprint arXiv:2601.14400, 2026

    Rafael Gómez-Lurbe and Armando Pérez. Pauli propagation for imaginary time evolution.arXiv preprint arXiv:2601.14400, 2026

    2026

  16. [41]

    Thermal state simulation with pauli and majorana propagation.arXiv preprint arXiv:2602.04878, 2026

    Manuel S Rudolph, Armando Angrisani, Andrew Wright, Iwo Sanderski, Ricard Puig, and Zoë Holmes. Thermal state simulation with pauli and majorana propagation.arXiv preprint arXiv:2602.04878, 2026

    2026

  17. [42]

    Dual role of low-weight pauli propagation: A flawed simulator but a powerful initializer for variational quantum algorithms.Phys

    Zong-Liang Li and Shi-Xin Zhang. Dual role of low-weight pauli propagation: A flawed simulator but a powerful initializer for variational quantum algorithms.Phys. Rev. Res., 8:013266, Mar 2026

    2026

  18. [43]

    Ollitrault, Cristian L

    Pauline J. Ollitrault, Cristian L. Cortes, Jérôme F. Gonthier, Robert M. Parrish, Dario Rocca, Gian-Luca Anselmetti, Matthias Degroote, Nikolaj Moll, Raffaele Santagati, and Michael Streif. Enhancing initial state overlap through orbital optimization for faster molecular elect...

    2024

  19. [44]

    Optimal polynomial based quantum eigenstate filtering with application to solving quantum linear systems.Quantum, 4:361, November 2020

    Lin Lin and Yu Tong. Optimal polynomial based quantum eigenstate filtering with application to solving quantum linear systems.Quantum, 4:361, November 2020

    2020

  20. [45]

    Kuzmin and Pietro Silvi

    Viacheslav V. Kuzmin and Pietro Silvi. Variational quantum state preparation via quantum data buses.Quantum, 4:290, July 2020

    2020

  21. [46]

    Variational quantum gibbs state preparation with a truncated taylor series

    Youle Wang, Guangxi Li, and Xin Wang. Variational quantum gibbs state preparation with a truncated taylor series. Phys. Rev. Appl., 16:054035, Nov 2021

    2021

  22. [47]

    Sornborger, and Patrick J

    Sumeet Khatri, Ryan LaRose, Alexander Poremba, Lukasz Cincio, Andrew T. Sornborger, and Patrick J. Coles. Quantum- assisted quantum compiling.Quantum, 3:140, May 2019

    2019

  23. [48]

    Benjamin

    Tyson Jones and Simon C. Benjamin. Robust quantum compilation and circuit optimisation via energy minimisation. Quantum, 6:628, January 2022

    2022

  24. [49]

    Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C

    M. Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C. Benjamin, Suguru Endo, Keisuke Fujii, Jarrod R. McClean, Kosuke Mitarai, Xiao Yuan, Lukasz Cincio, and Patrick J. Coles. Variational quantum algorithms.Nature Reviews Physics, 3(9):625–644, Sep 2021

    2021

  25. [50]

    Love, Alán Aspuru-Guzik, and Jeremy L

    Alberto Peruzzo, Jarrod McClean, Peter Shadbolt, Man-Hong Yung, Xiao-Qi Zhou, Peter J. Love, Alán Aspuru-Guzik, and Jeremy L. O’Brien. A variational eigenvalue solver on a photonic quantum processor.Nature Communications, 5(1):4213, Jul 2014

    2014

  26. [51]

    Synergistic pretraining of parametrized quantum circuits via tensor networks.Nature Communications, 14(1):8367, 2023

    Manuel S Rudolph, Jacob Miller, Danial Motlagh, Jing Chen, Atithi Acharya, and Alejandro Perdomo-Ortiz. Synergistic pretraining of parametrized quantum circuits via tensor networks.Nature Communications, 14(1):8367, 2023

    2023

  27. [52]

    Okada, Keita Osaki, Kosuke Mitarai, and Keisuke Fujii

    Ken N. Okada, Keita Osaki, Kosuke Mitarai, and Keisuke Fujii. Classically optimized variational quantum eigensolver with applications to topological phases.Phys. Rev. Res., 5:043217, Dec 2023

    2023

  28. [53]

    Pre-optimizing variational quantum eigensolvers with tensor networks

    Abid Khan, Bryan K Clark, and Norm M Tubman. Pre-optimizing variational quantum eigensolvers with tensor networks. arXiv preprint arXiv:2310.12965, 2023. 12

    2023 arXiv

  29. [54]

    Parameterized quantum circuits as machine learning models.Quantum Science and Technology, 4(4):043001, nov 2019

    Marcello Benedetti, Erika Lloyd, Stefan Sack, and Mattia Fiorentini. Parameterized quantum circuits as machine learning models.Quantum Science and Technology, 4(4):043001, nov 2019

    2019

  30. [55]

    Learning to learn with quantum neural networks via classical neural networks.arXiv preprint arXiv:1907.05415, 2019

    Guillaume Verdon, Michael Broughton, Jarrod R McClean, Kevin J Sung, Ryan Babbush, Zhang Jiang, Hartmut Neven, and Masoud Mohseni. Learning to learn with quantum neural networks via classical neural networks.arXiv preprint arXiv:1907.05415, 2019

    1907 arXiv

  31. [56]

    Approximate amplitude encoding in shallow parameterized quantum circuits and its application to financial market indicators.Phys

    Kouhei Nakaji, Shumpei Uno, Yohichi Suzuki, Rudy Raymond, Tamiya Onodera, Tomoki Tanaka, Hiroyuki Tezuka, Naoki Mitsuda, and Naoki Yamamoto. Approximate amplitude encoding in shallow parameterized quantum circuits and its application to financial market indicators.Phys. Rev. R...

    2022

  32. [57]

    Approximate complex amplitude encoding algorithm and its application to data classification problems.Phys

    Naoki Mitsuda, Tatsuhiro Ichimura, Kouhei Nakaji, Yohichi Suzuki, Tomoki Tanaka, Rudy Raymond, Hiroyuki Tezuka, Tamiya Onodera, and Naoki Yamamoto. Approximate complex amplitude encoding algorithm and its application to data classification problems.Phys. Rev. A, 109:052423, May 2024

    2024

  33. [58]

    AQER: A scalable and efficient data loader for digital quantum computers

    Kaining Zhang, Xinbiao Wang, Yuxuan Du, Min-Hsiu Hsieh, and Dacheng Tao. AQER: A scalable and efficient data loader for digital quantum computers. InThe Fourteenth International Conference on Learning Representations, 2026

    2026

  34. [59]

    Simulating physics with computers.International Journal of Theoretical Physics, 21(6/7), 1982

    Richard P Feynman. Simulating physics with computers.International Journal of Theoretical Physics, 21(6/7), 1982

    1982

  35. [60]

    Universal quantum simulators.Science, 273(5278):1073–1078, 1996

    Seth Lloyd. Universal quantum simulators.Science, 273(5278):1073–1078, 1996

    1996

  36. [61]

    Quantum computational networks.Proceedings of the Royal Society of London

    David Elieser Deutsch. Quantum computational networks.Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences, 425(1868):73–90, 09 1989

    1989

  37. [62]

    Helstrom

    Carl W. Helstrom. Quantum detection and estimation theory.Journal of Statistical Physics, 1(2):231–252, Jun 1969

    1969

  38. [63]

    P. J. J. O’Malley, R. Babbush, I. D. Kivlichan, J. Romero, J. R. McClean, R. Barends, J. Kelly, P. Roushan, A. Tranter, N. Ding, B. Campbell, Y. Chen, Z. Chen, B. Chiaro, A. Dunsworth, A. G. Fowler, E. Jeffrey, E. Lucero, A. Megrant, J. Y. Mutus, M. Neeley, C. Neill, C. Quinta...

    2016

  39. [64]

    Chow, and Jay M

    Abhinav Kandala, Antonio Mezzacapo, Kristan Temme, Maika Takita, Markus Brink, Jerry M. Chow, and Jay M. Gambetta. Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets.Nature, 549(7671):242–246, Sep 2017

    2017

  40. [65]

    Symmetric pruning in quantum neural networks

    Xinbiao Wang, Junyu Liu, Tongliang Liu, Yong Luo, Yuxuan Du, and Dacheng Tao. Symmetric pruning in quantum neural networks. InInternational Conference on Learning Representations, 2023

    2023

  41. [66]

    Mitarai, M

    K. Mitarai, M. Negoro, M. Kitagawa, and K. Fujii. Quantum circuit learning.Phys. Rev. A, 98:032309, Sep 2018

    2018

  42. [67]

    Córcoles, Kristan Temme, Aram W

    Vojtěch Havlíček, Antonio D. Córcoles, Kristan Temme, Aram W. Harrow, Abhinav Kandala, Jerry M. Chow, and Jay M. Gambetta. Supervised learning with quantum-enhanced feature spaces.Nature, 567(7747):209–212, Mar 2019

    2019

  43. [68]

    Svore, and Nathan Wiebe

    Maria Schuld, Alex Bocharov, Krysta M. Svore, and Nathan Wiebe. Circuit-centric quantum classifiers.Phys. Rev. A, 101:032308, Mar 2020

    2020

  44. [69]

    Offline quantum reinforcement learning in a conservative manner

    Zhihao Cheng, Kaining Zhang, Li Shen, and Dacheng Tao. Offline quantum reinforcement learning in a conservative manner. InProceedings of the AAAI conference on artificial intelligence, volume 37, pages 7148–7156, 2023

    2023

  45. [70]

    Quantum imitation learning.IEEE Transactions on Neural Networks and Learning Systems, 35(10):14190–14204, 2024

    Zhihao Cheng, Kaining Zhang, Li Shen, and Dacheng Tao. Quantum imitation learning.IEEE Transactions on Neural Networks and Learning Systems, 35(10):14190–14204, 2024

    2024

  46. [71]

    Artificial Intelligence (R0)

    Yuxuan Du, Xinbiao Wang, Naixu Guo, Zhan Yu, Yang Qian, Kaining Zhang, Min-Hsiu Hsieh, Patrick Rebentrost, and Dacheng Tao.A Gentle Introduction to Quantum Machine Learning. Artificial Intelligence (R0). Springer Nature Singapore, 2025

    2025

  47. [72]

    Pennylane: Automatic differentiation of hybrid quantum-classical computations.arXiv preprint arXiv:1811.04968, 2018

    Ville Bergholm, Josh Izaac, Maria Schuld, Christian Gogolin, Shahnawaz Ahmed, Vishnu Ajith, M Sohaib Alam, Guillermo Alonso-Linaje, Bharath AkashNarayanan, Ali Asadi, et al. Pennylane: Automatic differentiation of hybrid quantum-classical computations.arXiv preprint arXiv:1811...

    2018 arXiv

  48. [73]

    Quantum computing with qiskit.arXiv preprint arXiv:2405.08810, 2024

    Ali Javadi-Abhari, Matthew Treinish, Kevin Krsulich, Christopher J Wood, Jake Lishman, Julien Gacon, Simon Martiel, Paul D Nation, Lev S Bishop, Andrew W Cross, et al. Quantum computing with qiskit.arXiv preprint arXiv:2405.08810, 2024

    2024 arXiv

  49. [74]

    Evaluating analytic gradients on quantum hardware.Phys

    Maria Schuld, Ville Bergholm, Christian Gogolin, Josh Izaac, and Nathan Killoran. Evaluating analytic gradients on quantum hardware.Phys. Rev. A, 99:032331, Mar 2019

    2019

  50. [75]

    Gradients of parameterized quantum gates using the parameter-shift rule and gate decomposition.arXiv preprint arXiv:1905.13311, 2019

    Gavin E Crooks. Gradients of parameterized quantum gates using the parameter-shift rule and gate decomposition.arXiv preprint arXiv:1905.13311, 2019

    1905 arXiv

  51. [76]

    General parameter-shift rules for quantum gradients

    David Wierichs, Josh Izaac, Cody Wang, and Cedric Yen-Yu Lin. General parameter-shift rules for quantum gradients. Quantum, 6:677, March 2022

    2022

  52. [77]

    The power of quantum neural networks.Nature Computational Science, 1(6):403–409, Jun 2021

    Amira Abbas, David Sutter, Christa Zoufal, Aurelien Lucchi, Alessio Figalli, and Stefan Woerner. The power of quantum neural networks.Nature Computational Science, 1(6):403–409, Jun 2021

    2021

  53. [78]

    Caro, Hsin-Yuan Huang, M

    Matthias C. Caro, Hsin-Yuan Huang, M. Cerezo, Kunal Sharma, Andrew Sornborger, Lukasz Cincio, and Patrick J. Coles. Generalization in quantum machine learning from few training data.Nature Communications, 13(1):4919, Aug 2022

    2022

  54. [79]

    Separable power of classical and quantum learning protocols through the lens of no-free-lunch theorem.arXiv preprint arXiv:2405.07226, 2024

    Xinbiao Wang, Yuxuan Du, Kecheng Liu, Yong Luo, Bo Du, and Dacheng Tao. Separable power of classical and quantum learning protocols through the lens of no-free-lunch theorem.arXiv preprint arXiv:2405.07226, 2024

    2024 arXiv

  55. [80]

    Can variational quantum algorithms demonstrate quantum advantages? time really matters.arXiv preprint arXiv:2307.04089, 2023

    Huan-Yu Liu, Zhao-Yun Chen, Tai-Ping Sun, Cheng Xue, Yu-Chun Wu, and Guo-Ping Guo. Can variational quantum algorithms demonstrate quantum advantages? time really matters.arXiv preprint arXiv:2307.04089, 2023

    2023 arXiv

  56. [81]

    The heisenberg representation of quantum computers.arXiv preprint quant-ph/9807006, 1998

    Daniel Gottesman. The heisenberg representation of quantum computers.arXiv preprint quant-ph/9807006, 1998

    1998 arXiv

  57. [82]

    Stabilizer states and clifford operations for systems of arbitrary dimensions and modular arithmetic.Phys

    Erik Hostens, Jeroen Dehaene, and Bart De Moor. Stabilizer states and clifford operations for systems of arbitrary dimensions and modular arithmetic.Phys. Rev. A, 71:042315, Apr 2005. 13

    2005

  58. [83]

    Classical surrogate simulation of quantum systems with lowesa.arXiv preprint arXiv:2308.09109, 2023

    Manuel S Rudolph, Enrico Fontana, Zoë Holmes, and Lukasz Cincio. Classical surrogate simulation of quantum systems with lowesa.arXiv preprint arXiv:2308.09109, 2023

    2023 arXiv

  59. [84]

    Nemkov, Evgeniy O

    Nikita A. Nemkov, Evgeniy O. Kiktenko, and Aleksey K. Fedorov. Fourier expansion in variational quantum algorithms. Phys. Rev. A, 108:032406, Sep 2023

    2023

  60. [85]

    Quantum algorithms for quantum chemistry and quantum materials science.Chemical reviews, 120(22):12685–12717, 2020

    Bela Bauer, Sergey Bravyi, Mario Motta, and Garnet Kin-Lic Chan. Quantum algorithms for quantum chemistry and quantum materials science.Chemical reviews, 120(22):12685–12717, 2020

    2020

  61. [86]

    Effect of data encoding on the expressive power of variational quantum-machine-learning models.Physical Review A, 103(3):032430, 2021

    Maria Schuld, Ryan Sweke, and Johannes Jakob Meyer. Effect of data encoding on the expressive power of variational quantum-machine-learning models.Physical Review A, 103(3):032430, 2021

    2021

  62. [87]

    Egger, Jakub Mareček, and Stefan Woerner

    Daniel J. Egger, Jakub Mareček, and Stefan Woerner. Warm-starting quantum optimization.Quantum, 5:479, June 2021

    2021

  63. [88]

    Warm starting variational quantum algorithms with near clifford circuits

    Yun-Fei Niu, Shuo Zhang, and Wan-Su Bao. Warm starting variational quantum algorithms with near clifford circuits. Electronics, 12(2), 2023

    2023

  64. [89]

    Mitigating barren plateaus with transfer-learning-inspired parameter initializations.New Journal of Physics, 25(1):013039, 2023

    Huan-Yu Liu, Tai-Ping Sun, Yu-Chun Wu, Yong-Jian Han, and Guo-Ping Guo. Mitigating barren plateaus with transfer-learning-inspired parameter initializations.New Journal of Physics, 25(1):013039, 2023

    2023

  65. [90]

    Baptiste Anselme Martin and Thomas Ayral. Pre-optimization of quantum circuits, barren plateaus and classical simulability: tensor networks to unlock the variational quantum eigensolver.arXiv preprint arXiv:2602.04676, 2026

    2026

  66. [91]

    The one-dimensional ising model with a transverse field.Annals of Physics, 57(1):79–90, 1970

    Pierre Pfeuty. The one-dimensional ising model with a transverse field.Annals of Physics, 57(1):79–90, 1970

    1970

  67. [92]

    Rudolph, Armando Angrisani, Tyson Jones, M

    Sacha Lerch, Ricard Puig, Manuel S. Rudolph, Armando Angrisani, Tyson Jones, M. Cerezo, Supanut Thanasilp, and Zoë Holmes. Efficient quantum-enhanced classical simulation for patches of quantum landscapes.PRX Quantum, 7:020359, Jun 2026

    2026

  68. [93]

    Efficient learning for linear properties of bounded-gate quantum circuits

    Yuxuan Du, Min-Hsiu Hsieh, and Dacheng Tao. Efficient learning for linear properties of bounded-gate quantum circuits. Nature Communications, 16(1):3790, Apr 2025

    2025

  69. [94]

    Sample-efficient quantum error mitigation via classical learning surrogates.arXiv preprint arXiv:2511.07092, 2025

    Wei-You Liao, Ge Yan, Yujin Song, Tian-Ci Tian, Wei-Ming Zhu, De-Tao Jiang, Yuxuan Du, and He-Liang Huang. Sample-efficient quantum error mitigation via classical learning surrogates.arXiv preprint arXiv:2511.07092, 2025

    2025

  70. [95]

    Demonstration of efficient predictive surrogates for large-scale quantum processors.Nature Communications, 17(1):4731, May 2026

    Wei-You Liao, Yuxuan Du, Xinbiao Wang, Tian-Ci Tian, Yong Luo, Bo Du, Dacheng Tao, and He-Liang Huang. Demonstration of efficient predictive surrogates for large-scale quantum processors.Nature Communications, 17(1):4731, May 2026

    2026

  71. [96]

    A practical guide to using pauli path simulators for utility-scale quantum experiments.arXiv preprint arXiv:2507.10771, 2025

    Hrant Gharibyan, Siddharth Hariprakash, Mohammed Zuhair Mullath, and Vincent P Su. A practical guide to using pauli path simulators for utility-scale quantum experiments.arXiv preprint arXiv:2507.10771, 2025

    2025 arXiv

  72. [97]

    Pauli propagation: Simulating quantum spin dynamics via operator complexity.arXiv preprint arXiv:2510.22311, 2025

    Yuguo Shao, Song Cheng, and Zhengwei Liu. Pauli propagation: Simulating quantum spin dynamics via operator complexity.arXiv preprint arXiv:2510.22311, 2025

    2025 arXiv

  73. [98]

    Mele, Manuel S

    Armando Angrisani, Antonio A. Mele, Manuel S. Rudolph, M. Cerezo, and Zoë Holmes. Simulating quantum circuits with arbitrary local noise using pauli propagation.PRX Quantum, 7:020313, Apr 2026

    2026

  74. [99]

    Escaping from the barren plateau via gaussian initializations in deep variational quantum circuits.Advances in Neural Information Processing Systems, 35:18612–18627, 2022

    Kaining Zhang, Liu Liu, Min-Hsiu Hsieh, and Dacheng Tao. Escaping from the barren plateau via gaussian initializations in deep variational quantum circuits.Advances in Neural Information Processing Systems, 35:18612–18627, 2022

    2022

  75. [100]

    Trainability enhancement of parameterized quantum circuits via reduced-domain parameter initialization.Phys

    Yabo Wang, Bo Qi, Chris Ferrie, and Daoyi Dong. Trainability enhancement of parameterized quantum circuits via reduced-domain parameter initialization.Phys. Rev. Appl., 22:054005, Nov 2024

    2024

  76. [101]

    A unifying account of warm start guarantees for patches of quantum landscapes.arXiv preprint arXiv:2502.07889, 2025

    Hela Mhiri, Ricard Puig, Sacha Lerch, Manuel S Rudolph, Thiparat Chotibut, Supanut Thanasilp, and Zoë Holmes. A unifying account of warm start guarantees for patches of quantum landscapes.arXiv preprint arXiv:2502.07889, 2025

    2025 arXiv

  77. [102]

    TITAN: A trajectory-informed technique for adaptive parameter freezing in large-scale VQE

    Yifeng Peng, Xinyi Li, Samuel Yen-Chi Chen, Kaining Zhang, Zhiding Liang, Ying Wang, and Yuxuan Du. TITAN: A trajectory-informed technique for adaptive parameter freezing in large-scale VQE. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025

    2025

  78. [103]

    McClean, Sergio Boixo, Vadim N

    Jarrod R. McClean, Sergio Boixo, Vadim N. Smelyanskiy, Ryan Babbush, and Hartmut Neven. Barren plateaus in quantum neural network training landscapes.Nature Communications, 9(1):4812, Nov 2018

    2018

  79. [104]

    Coles, Lukasz Cincio, Jarrod R

    Martín Larocca, Supanut Thanasilp, Samson Wang, Kunal Sharma, Jacob Biamonte, Patrick J. Coles, Lukasz Cincio, Jarrod R. McClean, Zoë Holmes, and M. Cerezo. Barren plateaus in variational quantum computing.Nature Reviews Physics, 7(4):174–189, Apr 2025

    2025

  80. [105]

    The curse of random quantum data

    Kaining Zhang, Junyu Liu, Liu Liu, Liang Jiang, Min-Hsiu Hsieh, and Dacheng Tao. The curse of random quantum data. arXiv preprint arXiv:2408.09937, 2024. 14 CONTENTS I. Introduction 1 II. Preliminary 2 III. Biased gradient estimation forTb-PBS4 IV.SPPS: a faithful simulator fo...

    2024 arXiv

  81. [106]

    Sequential path sampling and importance reweighting 25

  82. [107]

    Choice of the smoothing parameter 26

  83. [108]

    Path automatic differentiation 26

  84. [109]

    Numerically stable PAD implementation 27

  85. [110]

    Single-sample variance for Gradient estimation 29

  86. [111]

    Convergence guarantee forSPPS-driven gradient descent optimization (Proof of Corollary 2) 34

    High-probability bounds for gradient estimation 32 F. Convergence guarantee forSPPS-driven gradient descent optimization (Proof of Corollary 2) 34

  87. [112]

    A variance-based first-order convergence guarantee for quantum optimization 34 15

  88. [113]

    Variance of the fullSPPSgradient estimator 37

  89. [114]

    Experimental settings and additional experimental results 40

    Convergence complexity ofSPPS-based SGD 39 G. Experimental settings and additional experimental results 40

  90. [115]

    Experimental settings 40

  91. [116]

    Additional VQE results 41

  92. [117]

    Subsequently, we briefly outline the implementation of thePBS method

    Additional state-preparation results 42 Appendix A: More preliminaries and related work In this appendix, we first introduce notations used in this work and the basics of quantum computing. Subsequently, we briefly outline the implementation of thePBS method. Finally, we provi...

  93. [118]

    For a positive integerN, we denote[N] ={1,...,N}

    Notations We summarize the notation used throughout this work. For a positive integerN, we denote[N] ={1,...,N} . Vectors are written in bold font, for example,aj denotes thej-th component of a vectora. The tensor product is denoted by ⊗, the conjugate transpose ofA byA†, and ...

  94. [119]

    A qubit is a two-level quantum-mechanical system described by a unit vector in the Hilbert spaceC2

    Basics of quantum computing Basics of quantum computation.The elementary unit of quantum computation is the qubit (or quantum bit), which is the quantum mechanical analog of a classical bit. A qubit is a two-level quantum-mechanical system described by a unit vector in the Hil...

  95. [120]

    Pauli path propagation in the Heisenberg picture LetPn ={sP :s∈{± 1},P∈{I,X,Y,Z} ⊗n} be the set ofn-qubit Pauli strings with an overall sign±1. Then, any Hermitian observable admits a Pauli expansion O= NOX k=1 ckOk, O k∈P n, ck >0.(A3) We consider a parameterized circuit cons...

  96. [121]

    PBSprovidesaHeisenberg-picture framework for estimating expectation values by expanding the evolved observable in the Pauli basis [20, 22–25, 27, 29]

    Related work Wereviewprior PBSmethodsthataremostrelevanttotruncation-based PBS(Tb-PBS). PBSprovidesaHeisenberg-picture framework for estimating expectation values by expanding the evolved observable in the Pauli basis [20, 22–25, 27, 29]. Since the number of propagation paths ...

  97. [122]

    The construction uses a single-qubit observable, while a Clifford fanout layer generates high-weight Pauli strings during Heisenberg propagation

    Explicit case with standard truncation rules We next provide an explicitn-qubit construction showing thatCT, FT, and WT can realize the value-gradient separation in Theorem 3. The construction uses a single-qubit observable, while a Clifford fanout layer generates high-weight ...

  98. [123]

    Sequential path sampling and importance reweighting We follow the notation introduced in Section A3. For a Pauli observableO∈P n, the Heisenberg-picture propagation gives f(θ) = Tr OU(θ)ρU(θ)† = X ω∈Ω Ψω(θ)Tr [Pω(O)ρ],(D1) whereω = (ω1,...,ω L)∈ Ω⊆{ 0,± 1}L denotes a legal pro...

  99. [124]

    (D5) controls the trade-off between magnitude-proportional sampling and exploration of derivative-sensitive branches

    Choice of the smoothing parameter The parametera in Eq. (D5) controls the trade-off between magnitude-proportional sampling and exploration of derivative-sensitive branches. In practice,SPPSsupports two choices ofa. Fixed smoothing.The simplest choice is to use a constant smoo...

  100. [125]

    Path automatic differentiation We now derive the PAD estimator used to obtain gradients from the sampled paths. For a fixed pathω, its contribution after importance reweighting as ˜hω(θ) = Ψω(θ) Pr(ω) Tr[Pω(O)ρ] = 1 Pr(ω) Tr[Pω(O)ρ] Y j∈A(ω) Ψωj(θj),(D14) whereA(ω) ={j:ω j̸= 0...

  101. [126]

    For example, ifωj = 1and |cosθ j|≈ 0, the path value˜hω(θ)can be very small while−tanθ j is very large

    Numerically stable PAD implementation The score form in Eq.(D18) is algebraically convenient but can be numerically unstable when|cosθ j| or|sinθ j| is close to zero. For example, ifωj = 1and |cosθ j|≈ 0, the path value˜hω(θ)can be very small while−tanθ j is very large. Their ...

  102. [127]

    To avoid using a fixed overly conservative budget, the implementation uses an adaptive absolute gradient-error proxy based on two independent macro-replicates

    Adaptive gradient-error proxy The number of samples required bySPPS varies across optimization steps and observable terms. To avoid using a fixed overly conservative budget, the implementation uses an adaptive absolute gradient-error proxy based on two independent macro-replic...

  103. [128]

    First, it constructs the local sampling probabilities in Eq.(D11) from the current parameters

    Implementation summary For each optimization step,SPPS proceeds as follows. First, it constructs the local sampling probabilities in Eq.(D11) from the current parameters. Second, for each Pauli term, it draws two independent groups of propagation paths and evaluates both value...

  104. [129]

    (D1) is defined as ˜gω(θ) :=∇θΨω(θ) Pr(ω) Tr [Pω(O)ρ].(E1) We denote itsk-th coordinate by˜gω,k(θ)

    Single-sample variance for Gradient estimation For a pathω∈Ω, the single-sample gradient estimator for the value functionf(θ)in Eq. (D1) is defined as ˜gω(θ) :=∇θΨω(θ) Pr(ω) Tr [Pω(O)ρ].(E1) We denote itsk-th coordinate by˜gω,k(θ). The empirical gradient estimator is then give...

  105. [130]

    We first recall the Bernstein inequality used below and then prove the formal gradient-estimation statement of Theorem 2

    High-probability bounds for gradient estimation We now combine the variance bound established in Section E1 with Bernstein’s inequality to obtain a high-probability error bound for the empirical gradient estimator. We first recall the Bernstein inequality used below and then p...

  106. [131]

    A variance-based first-order convergence guarantee for quantum optimization We first derive a generic first-order convergence guarantee for SGD when optimizing the objective function in Eq. (F1). This result is stated in terms of a variance bound on the stochastic gradient, an...

  107. [132]

    Specifically, we suppose thatO=PNO m=1cmOm, where eachOm∈P n

    Variance of the fullSPPSgradient estimator We now bound the variance of the fullSPPS gradient estimator for a general observable expressed in the Pauli basis. Specifically, we suppose thatO=PNO m=1cmOm, where eachOm∈P n. For eachm∈[N O], define f(m)(θ) := Tr OmU(θ)ρU(θ)† .(F28...

  108. [133]

    The resulting guarantee is stated explicitly in Theorem 6

    Convergence complexity ofSPPS-based SGD We are now ready to combine the generic convergence guarantee in Lemma 4 with the variance bound for the fullSPPS gradient estimator in Lemma 5 to derive the convergence complexity ofSPPS-based SGD. The resulting guarantee is stated expl...

  109. [134]

    The variational circuit is initialized from the plus state by default and uses a one-dimensional hardware-efficient ansatz

    Experimental settings VQE settings.The Hamiltonian is H=−J n−1X i=1 ZiZi+1−g nX i=1 Xi,(G1) with J = 1.0and g = 1.0unless otherwise specified. The variational circuit is initialized from the plus state by default and uses a one-dimensional hardware-efficient ansatz. Each layer...

  110. [135]

    5 comparesSPPS with the tensor-network baseline on theL = 6TFIM VQE benchmark

    Additional VQE results SPPS remains competitive against tensor-network simulation.Fig. 5 comparesSPPS with the tensor-network baseline on theL = 6TFIM VQE benchmark. Acrossn = 20to100qubits, SPPS attains a favorable error-runtime trade-off. The advantage is especially clear at...

  111. [136]

    Additional state-preparation results SPPS prepares quantum encoding circuits for MNIST amplitude encoding.We finally evaluate whether SPPS can be used beyond VQE and QML pre-training by optimizing state-preparation circuits. Fig. 8 summarizes the results. Figures (a) and (b) s...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.