pith. machine review for the scientific record. sign in

arxiv: 2604.25760 · v1 · submitted 2026-04-28 · 🪐 quant-ph

Recognition: unknown

Beyond Single Trajectories: Optimal Control and Jordan-Lie Algebra in Hybrid Quantum Walks for Combinatorial Optimization

Tianen Chen , Yun Shang
Authors on Pith no claims yet

Pith reviewed 2026-05-07 16:31 UTC · model grok-4.3

classification 🪐 quant-ph
keywords hybrid quantum walkQAOAcombinatorial optimizationJordan-Lie algebraoptimal controlPontryagin minimum principleMax-CutMaximum Independent Set
0
0 comments X

The pith

A hybrid quantum walk ansatz using dynamical coin operators outperforms QAOA by coherently superposing multiple evolution paths and generating a larger Jordan-Lie algebra.

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

The paper introduces a hybrid quantum walk framework that superposes multiple Hamiltonian-driven paths coherently in each layer through a dynamical coin operator, treating QAOA as the special case with a fixed Pauli-X coin. Using optimal control theory via Pontryagin's minimum principle, it finds that the optimal coin is generally time-dependent rather than constant, and algebraic analysis shows this creates a strictly larger Jordan-Lie algebra whose unique negative Jordan products correlate with performance gains. Experiments on Max-Cut and Maximum Independent Set instances confirm faster convergence, higher solution quality, and greater robustness than standard QAOA. Readers should care because this suggests a new paradigm for variational quantum algorithms that leverages path superposition to enhance expressivity without necessarily increasing circuit complexity.

Core claim

The HQW ansatz superposes multiple trajectories coherently within each circuit layer via a dynamical coin operator derived from Pontryagin's minimum principle. This generates a strictly larger Jordan-Lie algebra than QAOA, with the negativity of the Jordan product providing an algebraic explanation for improved optimization performance. Numerical results on combinatorial problems establish that HQW achieves superior convergence speed, accuracy, and robustness.

What carries the argument

The dynamical coin operator, obtained through optimal control, that enables coherent superposition of multiple paths and enlarges the generated Jordan-Lie algebra for greater variational expressivity.

If this is right

  • QAOA is recovered as the special case with a static coin, so improvements apply broadly to variational optimization.
  • The optimal coin operator differs from constant gates, implying that time-varying controls can enhance algorithm performance.
  • Jordan product negativity in the dynamical Lie algebra is tied to better convergence and solution accuracy.
  • Path-superposition via optimal control offers a systematic way to design more expressive quantum optimization algorithms.

Where Pith is reading between the lines

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

  • This framework might extend to other quantum algorithms by incorporating similar superposition of control trajectories to boost expressivity.
  • Reduced layer counts could be possible due to higher expressivity per layer, lowering hardware requirements.
  • Further analysis could link the algebraic structure directly to approximation guarantees for specific problems like Max-Cut.
  • The robustness advantage suggests potential resilience to certain noise types in NISQ devices.

Load-bearing premise

That the optimal dynamical coin operator can be realized on hardware with manageable circuit depth and that the benefits from path superposition are not canceled out by additional noise or implementation overhead.

What would settle it

A direct comparison on a Max-Cut problem where HQW fails to show faster convergence or higher accuracy than QAOA with the same number of layers, or where the Jordan-Lie algebra size is not larger.

Figures

Figures reproduced from arXiv: 2604.25760 by Tianen Chen, Yun Shang.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic of a view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Schematic of a view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. The function graph of view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Geometric interpretation of the Jordan product’s role. (Left) When view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Correlation between Jordan negativity view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Distribution of HQW’s relative improvement over view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Scatter plot of average 1 view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Scatter plot of average 1 view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. The 8-vertex graph for the Max-Cut problem. view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11. The 8-vertex graph for the Maximum Independent view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12. Experimental results on an 8-vertex Max-Cut instance. view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13. Experimental results on an 8-vertex MIS instance. view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14. The 12-vertex graph for the Max-Cut problem. view at source ↗
Figure 16
Figure 16. Figure 16: FIG. 16. Performance comparison on a 12-vertex Max-Cut view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17. Performance comparison on a 10-vertex MIS graph. view at source ↗
Figure 18
Figure 18. Figure 18: FIG. 18. A scatter plot of best 1 view at source ↗
Figure 19
Figure 19. Figure 19: FIG. 19. Distribution of HQW’s relative improvement over view at source ↗
read the original abstract

The Quantum Approximate Optimization Algorithm (QAOA) follows a single, fixed evolution path, overlooking the potential computational advantage of coherently superposing multiple trajectories. Here we overcome this limitation with a hybrid quantum walk (HQW) ansatz that super poses multiple Hamiltonian-driven paths coherently within each circuit layer via a dynamical coin operator. QAOA emerges as a special case of this framework with a static Pauli-X coin. Using Pontryagin's minimum principle, we derive the optimal form of the coin operator, demonstrating that it generally differs from a constant gate. A dynamical Lie algebra analysis reveals that HQW generates a strictly larger Jordan-Lie algebra, providing an algebraic foundation for its enhanced expressivity. Especially, we reveal the connection between the unique Jordan product negativity in HQW's DLA and its performance advantages. Numerical experiments on Max-Cut and Maximum Independent Set problems show that HQW systematically outperforms QAOA in convergence speed, solution accuracy, and robustness. Our work establishes a path-superposition paradigm for quantum optimization, combining optimal control theory with algebraic structure to guide the design of advanced quantum algorithms.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces a hybrid quantum walk (HQW) ansatz for combinatorial optimization that extends QAOA by coherently superposing multiple Hamiltonian-driven trajectories per layer via a dynamical coin operator. QAOA is recovered as the special case of a static Pauli-X coin. The optimal coin is derived via Pontryagin's minimum principle and shown to differ from a constant gate in general. A dynamical Lie algebra analysis establishes that HQW spans a strictly larger Jordan-Lie algebra than QAOA, with a connection drawn between unique Jordan-product negativity and performance gains. Numerical experiments on Max-Cut and Maximum Independent Set problems report that HQW outperforms QAOA in convergence speed, solution accuracy, and robustness.

Significance. If the reported advantages survive under fixed total gate budgets, the work would meaningfully advance variational quantum optimization by combining optimal-control-derived ansatze with Lie-algebraic diagnostics. The formal demonstration of an enlarged Jordan-Lie algebra and the proposed link to Jordan negativity supply a concrete theoretical handle for designing more expressive circuits. The numerical results on two standard combinatorial problems provide initial evidence, but the absence of resource-equivalent comparisons limits immediate claims of practical superiority.

major comments (2)
  1. [Numerical Experiments] Numerical Experiments section: The central claim that HQW systematically outperforms QAOA rests on layer-count comparisons. The manuscript states that the Pontryagin-derived dynamical coin 'generally differs from a constant gate,' yet provides no decomposition into elementary gates, no total two-qubit gate counts, and no ablation that holds total gate budget fixed rather than layer number. Without this accounting, the observed gains in convergence, accuracy, and robustness could arise from unequal computational resources rather than the larger Jordan-Lie algebra or Jordan-product negativity.
  2. [Dynamical Lie Algebra Analysis] Dynamical Lie Algebra Analysis section: The paper correctly shows that HQW generates a strictly larger Jordan-Lie algebra. However, it does not quantify the additional reachable expressivity within hardware depth limits nor demonstrate that the extra generators are actually utilized by the optimized circuits in the numerical experiments. A direct, quantitative bridge between the algebraic enlargement and the reported performance metrics is needed to support the claim that the larger algebra provides the 'algebraic foundation' for the advantages.
minor comments (2)
  1. [Introduction] The abstract and introduction refer to 'Jordan product negativity' as uniquely connected to performance; a brief explicit definition or equation reference in the main text would aid readers unfamiliar with the signed Jordan product.
  2. [Numerical Experiments] Figure captions for the numerical results should state the number of independent runs, the optimizer used, and whether error bars represent standard deviation or standard error.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below, providing clarifications on the current manuscript and committing to targeted revisions that directly strengthen the evidence for our claims.

read point-by-point responses

  1. Referee: [Numerical Experiments] Numerical Experiments section: The central claim that HQW systematically outperforms QAOA rests on layer-count comparisons. The manuscript states that the Pontryagin-derived dynamical coin 'generally differs from a constant gate,' yet provides no decomposition into elementary gates, no total two-qubit gate counts, and no ablation that holds total gate budget fixed rather than layer number. Without this accounting, the observed gains in convergence, accuracy, and robustness could arise from unequal computational resources rather than the larger Jordan-Lie algebra or Jordan-product negativity.

    Authors: We agree that resource-equivalent comparisons are essential to isolate the contribution of the enlarged Jordan-Lie algebra. Although layer count is the conventional metric in the QAOA literature and the dynamical coin is a fixed-size unitary (decomposable into a constant number of elementary gates per layer), the manuscript does not explicitly report total two-qubit gate counts or perform fixed-budget ablations. In the revised version we will add: (i) an explicit elementary-gate decomposition of the Pontryagin-derived coin, (ii) the resulting total two-qubit gate counts for HQW versus QAOA at matched layer depths, and (iii) a new ablation study that holds the total gate budget fixed while varying the coin dynamics. These additions will demonstrate that the reported advantages in convergence speed, accuracy, and robustness persist under equal resource constraints. revision: yes

  2. Referee: [Dynamical Lie Algebra Analysis] Dynamical Lie Algebra Analysis section: The paper correctly shows that HQW generates a strictly larger Jordan-Lie algebra. However, it does not quantify the additional reachable expressivity within hardware depth limits nor demonstrate that the extra generators are actually utilized by the optimized circuits in the numerical experiments. A direct, quantitative bridge between the algebraic enlargement and the reported performance metrics is needed to support the claim that the larger algebra provides the 'algebraic foundation' for the advantages.

    Authors: The DLA analysis rigorously establishes strict inclusion of the QAOA algebra inside the HQW algebra and links the unique Jordan-product negativity to improved optimization landscapes. However, the current manuscript does not quantify the dimension of the reachable subalgebra under realistic hardware-depth constraints nor measure the participation of the additional generators in the numerically optimized circuits. In the revision we will insert: (1) the explicit dimension of the full DLA versus the hardware-constrained reachable algebra, (2) a quantitative analysis (e.g., participation ratios and gradient-flow diagnostics) showing that the extra generators are actively utilized by the optimized HQW circuits on the Max-Cut and MIS instances, and (3) a concise discussion connecting these algebraic metrics to the observed gains in convergence and solution quality. This will furnish the direct quantitative bridge requested. revision: yes

Circularity Check

0 steps flagged

No circularity: derivations rely on external optimal control and independent algebraic analysis

full rationale

The paper's core derivations do not reduce to their own inputs by construction. The optimal dynamical coin is obtained via Pontryagin's minimum principle (an external optimal-control theorem), not by fitting or self-definition within the HQW ansatz. The Jordan-Lie algebra comparison is a formal algebraic inclusion proof showing a strictly larger DLA for the hybrid walk versus QAOA; it contains no fitted parameters, no self-referential predictions, and no load-bearing self-citations. QAOA appears only as the special case of a static Pauli-X coin, which is a direct substitution rather than a circular prediction. Numerical performance claims on Max-Cut and MIS are empirical outputs, not re-statements of the algebraic or control derivations. No ansatz is smuggled via prior author work, no uniqueness theorem is imported from the same authors, and no known empirical pattern is merely renamed. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard quantum control theory and Lie algebra concepts without introducing new free parameters or invented entities in the abstract.

axioms (1)
  • domain assumption Pontryagin's minimum principle can be applied to derive the optimal dynamical coin operator for the hybrid quantum walk
    Invoked to obtain the optimal form of the coin operator that differs from a constant gate.

pith-pipeline@v0.9.0 · 5493 in / 1199 out tokens · 53799 ms · 2026-05-07T16:31:27.267124+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

56 extracted references · 11 canonical work pages · 2 internal anchors

  1. [1]

    3.k ∗(tf) =∇ xϕ(x∗(tf), tf)

    ˙k ∗ (t) =−∇ xH=−∇ xg−k ∗(t)· ∇ xf. 3.k ∗(tf) =∇ xϕ(x∗(tf), tf). 4.H(x ∗(t),k ∗(t),v ∗(t)) = min v∈U H(x ∗(t),k ∗(t),v) for a.e.t∈[t 0, tf]. These equations form a two-point boundary value prob- lem, and the PMP often reduces the optimal control problem to solving forv ∗(t) that minimizes the control Hamiltonian. D. Dynamical Lie algebra In the context of...

  2. [2]

    | ˙ψ(t)⟩=−iH(t)|ψ(t)⟩,|ψ(t 0)⟩=|ψ(0)⟩

    Dynamical equation. | ˙ψ(t)⟩=−iH(t)|ψ(t)⟩,|ψ(t 0)⟩=|ψ(0)⟩. The man- ifoldMis the unit sphere inC n

  3. [3]

    Control input.v(t) = (u(t),n(t))∈ U={0, 1 2 ,1}× {x∈R 3,||x|| 2 = 1} ⊂R 4

  4. [4]

    Objective function. In order to be coincide with ⟨ψf |Hc|ψf ⟩, now the objective function is J=⟨ψ(t f)|I⊗H c|ψ(tf)⟩.(20) Thus, the control Hamiltonian H=y 1(t)Φ|1⟩⟨1|⊗Hc(t) +y 2(t)Φ|0⟩⟨0|⊗Hb(t) +y3(t)Φ(nx(t)X+n y(t)Y+n z(t)Z)⊗I (t),(21) where Φ A(t) =i⟨k(t)|A|ψ(t)⟩+ c.c., and the adjoint state|k(t)⟩satisfies the equation |˙k(t)⟩=−iH(t)|k(t)⟩,|k(t f)⟩=I⊗H ...

  5. [5]

    2.n 2 x(t) +n 2 y(t) +n 2 z(t) = 1

    ∂L ∂nx =−y 3(t)ΦX⊗I (t) + 2µnx = 0, ∂L ∂ny =−y 3(t)ΦY⊗I (t) + 2µny = 0, ∂L ∂nz =−y 3(t)ΦZ⊗I (t) + 2µnz = 0. 2.n 2 x(t) +n 2 y(t) +n 2 z(t) = 1. Whenn 2 x(t) +n 2 y(t) +n 2 z(t) = 1, andn x = y3(t)ΦX⊗I (t) 2µ , ny = y3(t)ΦY⊗I (t) 2µ andn z = y3(t)ΦZ⊗I (t) 2µ , we have 1 2µ = 1 |y3(t)| p ΦX⊗I (t)2 + ΦY⊗I (t)2 + ΦZ⊗I (t)2 .(24) Thus, when the control paramet...

  6. [6]

    [A, B◦C] = [A, B]◦C+B◦[A, C] and [A,[B, C]] = [[A, B], C] + [B,[A, C]] for allA, B, C∈ L

  7. [7]

    According to the definition of the Jordan-Lie algebra, we next present a lemma to describe an algebraL Q

    (A◦B)◦C−A◦(B◦C) =h 2[[A, C], B] for some h2 ∈R. According to the definition of the Jordan-Lie algebra, we next present a lemma to describe an algebraL Q. Lemma 1.L Q is the real vector space generated lin- early by{iH c, iHb, iIp}equipped with two bilinear maps i{·,·}and [·,·] :L Q × LQ → LQ, where{·,·}and [·,·] are the anti-commutator and the commutator,...

  8. [8]

    Problem Hamiltonians For a graphG= (V, E), the Max-Cut problem is en- coded using the standard Hamiltonian Hc = X (i,j)∈E I−Z iZj 2 ,(B1) with the mixer HamiltonianH b =P i Xi. The Maximum Independent Set problem is encoded as Hc =− X i ni +λ X (i,j)∈E ninj,(B2) wheren i = 1 2(I−Z i) represents the occupation num- ber operator for vertexi, indicating whet...

  9. [9]

    After Hadamard initialization, each layerlapplies: U(γ l, βl) = exp(−iγlHc) exp(−iβlHb) (B4) whereγ l andβ l are variational parameters optimized classically

    Circuit Implementations The classic QAOA circuit uses onlynsystem qubits. After Hadamard initialization, each layerlapplies: U(γ l, βl) = exp(−iγlHc) exp(−iβlHb) (B4) whereγ l andβ l are variational parameters optimized classically. The hybrid quantum walk circuit employsn+1 qubits: nsystem qubits representing graph vertices and 1 coin qubit controlling t...

  10. [10]

    Apply aU 3(θs,2, θs,3, θs,4) gate on the coin qubit, where U3(θ, ϕ, δ) = cos(θ/2)−exp(iδ) sin(θ/2) exp(iϕ) sin(θ/2) exp(i(ϕ+δ)) cos(θ/2) . (B5)

  11. [11]

    If the coin qubit is in state|0⟩, apply exp(−iθs,1Hb) to system qubits

    If the coin qubit is in state|1⟩, apply exp(−iθ s,0Hc) to system qubits. If the coin qubit is in state|0⟩, apply exp(−iθs,1Hb) to system qubits. The controlled evolution is implemented using Penny- Lane’s ctrl operation with conditional application of Ap- proxTimeEvolution. Each layer of QAOA adds two evo- lution processes corresponding to the two Hamilto...

  12. [12]

    Classic QAOA: The standard QAOA ansatz Ub(βp)Uc(γp)· · ·U b(β1)Uc(γ1)

  13. [13]

    Hybrid Quantum Walk (HQW): The HQW opera- tor W(γ k, βk) =e −i|1⟩⟨1|⊗Hcγk e−i|0⟩⟨0|⊗Hbβk(Ck ⊗I), withpsteps

  14. [14]

    QAOA Variant 1: A QAOA circuit where parame- tersγ k (forH c) are taken from the odd-step HQW parameters andβ k (forH b) from the even-step HQW parameters

  15. [15]

    For HQW, the final expectation value is⟨−I⊗H c⟩for the Max-Cut problem and⟨I⊗H c⟩for the Maximum Independent Set problem

    QAOA Variant 2: A QAOA circuit withγ k from even-step andβ k from odd-step HQW parameters. For HQW, the final expectation value is⟨−I⊗H c⟩for the Max-Cut problem and⟨I⊗H c⟩for the Maximum Independent Set problem. All circuits were optimized to minimize this expectation

  16. [16]

    equal performance

    Detailed Discussion of best-performance Improvements on 50 Graphs Here we provide additional details for the best- performance comparison on 50 Graphs. As noted in the main text, the experimental protocol is identical to that described for the best-performance comparison: QAOA depthp= 2, HQW steps 2p= 4, 100 random initial- izations, 300 Adam steps with l...

  17. [17]

    A Quantum Approximate Optimization Algorithm

    E. Farhi, J. Goldstone, and S. Gutmann, arXiv preprint arXiv:1411.4028 (2014)

    work page internal anchor Pith review arXiv 2014

  18. [18]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, arXiv preprint arXiv:1412.6062 (2015)

    work page arXiv 2015

  19. [19]

    Peruzzo, J

    A. Peruzzo, J. McClean, P. Shadbolt, M.-H. Yung, X.-Q. Zhou, P. J. Love, A. Aspuru-Guzik, and J. L. O’brien, Nature communications5, 4213 (2014)

    2014

  20. [20]

    G. E. Crooks, arXiv preprint arXiv:1811.08419 (2018)

    work page arXiv 2018

  21. [21]

    Choi and J

    J. Choi and J. Kim, in2019 international conference on information and communication technology convergence (ICTC)(IEEE, 2019) pp. 138–142

    2019

  22. [22]

    Zhou, S.-T

    L. Zhou, S.-T. Wang, S. Choi, H. Pichler, and M. D. Lukin, Physical Review X10, 021067 (2020)

    2020

  23. [23]

    Herrman, P

    R. Herrman, P. C. Lotshaw, J. Ostrowski, T. S. Humble, and G. Siopsis, Scientific Reports12, 6781 (2022)

    2022

  24. [24]

    Bravyi, A

    S. Bravyi, A. Kliesch, R. Koenig, and E. Tang, Physical review letters125, 260505 (2020)

    2020

  25. [25]

    L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamkrelidze, and E. F. Mishchenko,The Mathematical Theory of Op- timal Processes(John Wiley and Sons, New York, 1962)

    1962

  26. [26]

    Z.-C. Yang, A. Rahmani, A. Shabani, H. Neven, and C. Chamon, Physical Review X7, 021027 (2017)

    2017

  27. [27]

    C. Lin, Y. Wang, G. Kolesov, and U. Kalabi´ c, Physical Review A100, 022327 (2019)

    2019

  28. [28]

    L. T. Brady, C. L. Baldwin, A. Bapat, Y. Kharkov, and A. V. Gorshkov, Physical Review Letters126, 070505 (2021)

    2021

  29. [29]

    Gerblich, T

    L. Gerblich, T. Dasanjh, H. Q. Wong, D. Ross, L. Novo, N. Chancellor, and V. Kendon, arXiv preprint arXiv:2407.06663 (2024)

    work page arXiv 2024

  30. [30]

    D. A. Meyer, Journal of Statistical Physics85, 551 (1996)

    1996

  31. [31]

    D. A. Meyer, Physics Letters A223, 337 (1996)

    1996

  32. [32]

    Aharonov, A

    D. Aharonov, A. Ambainis, J. Kempe, and U. Vazirani, inProceedings of the thirty-third annual ACM symposium on Theory of computing(2001) pp. 50–59

    2001

  33. [33]

    Ambainis, E

    A. Ambainis, E. Bach, A. Nayak, A. Vishwanath, and J. Watrous, inProceedings of the thirty-third annual ACM symposium on Theory of computing(2001) pp. 37– 49

    2001

  34. [34]

    A. M. Childs, R. Cleve, E. Deotto, E. Farhi, S. Gutmann, and D. A. Spielman, inProceedings of the thirty-fifth an- nual ACM symposium on Theory of computing(2003) pp. 59–68

    2003

  35. [35]

    Chen and Y

    T. Chen and Y. Shang, npj Quantum Information12, 21 (2026)

    2026

  36. [36]

    Albertini and D

    F. Albertini and D. D’Alessandro, inProceedings of the 40th IEEE Conference on Decision and Control (Cat. No. 01CH37228), Vol. 2 (IEEE, 2001) pp. 1589–1594

    2001

  37. [37]

    Albertini and D

    F. Albertini and D. D’Alessandro, Systems & Control Letters151, 104913 (2021)

    2021

  38. [38]

    D’Alessandro and J

    D. D’Alessandro and J. T. Hartwig, Journal of Dynamical and Control Systems27, 1 (2021)

    2021

  39. [39]

    Wiersema, E

    R. Wiersema, E. K¨ okc¨ u, A. F. Kemper, and B. N. Bakalov, npj Quantum Information10, 110 (2024)

    2024

  40. [40]

    N. Diaz, D. Garc´ ıa-Mart´ ın, S. Kazi, M. Larocca, and M. Cerezo, arXiv preprint arXiv:2310.11505 (2023)

    work page arXiv 2023

  41. [41]

    Allcock, M

    J. Allcock, M. Santha, P. Yuan, and S. Zhang, arXiv preprint arXiv:2407.12587 (2024)

    work page arXiv 2024

  42. [42]

    Ragone, B

    M. Ragone, B. N. Bakalov, F. Sauvage, A. F. Kemper, C. Ortiz Marrero, M. Larocca, and M. Cerezo, Nature Communications15, 7172 (2024)

    2024

  43. [43]

    N. P. Landsman,Mathematical topics between classical and quantum mechanics(Springer Science & Business Media, 2012)

    2012

  44. [44]

    Quantum Computation by Adiabatic Evolution

    E. Farhi, J. Goldstone, S. Gutmann, and M. Sipser, arXiv preprint quant-ph/0001106 (2000)

    work page Pith review arXiv 2000

  45. [45]

    Albash and D

    T. Albash and D. A. Lidar, Reviews of Modern Physics 90, 015002 (2018)

    2018

  46. [46]

    Farhi, J

    E. Farhi, J. Goldstone, and S. Gutmann, arXiv preprint quant-ph/0208135 (2002)

    work page arXiv 2002

  47. [47]

    Seki and H

    Y. Seki and H. Nishimori, Physical Review E—Statistical, Nonlinear, and Soft Matter Physics 85, 051112 (2012)

    2012

  48. [48]

    Albash, Physical Review A99, 042334 (2019)

    T. Albash, Physical Review A99, 042334 (2019)

    2019

  49. [49]

    Ghosh, L

    R. Ghosh, L. A. Nutricati, N. Feinstein, P. Warburton, and S. Bose, arXiv preprint arXiv:2409.13029 (2024)

    work page arXiv 2024

  50. [50]

    S. Sim, P. D. Johnson, and A. Aspuru-Guzik, Advanced Quantum Technologies2, 1900070 (2019)

    2019

  51. [51]

    Guo and S

    Y. Guo and S. Luo, Physical Review A110, 062206 (2024)

    2024

  52. [52]

    PennyLane: Automatic differentiation of hybrid quantum-classical computations

    V. Bergholm, J. Izaac, M. Schuld, C. Gogolin, S. Ahmed, V. Ajith, M. S. Alam, G. Alonso-Linaje, B. AkashNarayanan, A. Asadi,et al., arXiv preprint arXiv:1811.04968 (2018)

    work page internal anchor Pith review arXiv 2018

  53. [53]

    T. E. Oliphantet al.,Guide to numpy, Vol. 1 (Trelgol Publishing USA, 2006)

    2006

  54. [54]

    J. D. Hunter, Computing in science & engineering9, 90 (2007)

    2007

  55. [55]

    McKinneyet al., scipy445, 51 (2010)

    W. McKinneyet al., scipy445, 51 (2010). 18

    2010

  56. [56]

    Chen and Y

    T. Chen and Y. Shang, 10.5281/zenodo.18983535 (2026). [41] D. P. Kingma and J. Ba, arXiv preprint arXiv:1412.6980 (2014)

    work page doi:10.5281/zenodo.18983535 2026