Pith. sign in

REVIEW 4 major objections 4 minor 48 references

An Exclusive-Sum-of-Products Pipeline for QAOA

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that encoding MIS constraints as exclusive-sum-of-products (ESOP) Boolean expressions before penalization yields QAOA approximation ratios up to 30.3% higher than standard QUBO penalty methods, with improvements on about 64

desk verdict The paper's central derivation of the ESOP constraint Hamiltonian has multiple load-bearing algebraic errors, so the reported 30% improvements are not measurements of the claimed pipeline. read the letter →

arxiv 2508.21686 v1 pith:IUKKFGMX submitted 2025-08-29 quant-ph

Matthew Brunet , Shilpi Shah , Mostafa Atallah , Anthony Wilkie , Rebekah Herrman This is my paper
classification quant-ph
keywords QAOAexclusive-sum-of-productsESOPmaximumindependentsetconstraintpenalizationapproximationratioHamiltonianencoding
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 proposes a new pipeline for incorporating constraints into the quantum approximate optimization algorithm (QAOA): rewrite each constraint as a Boolean expression, convert it to exclusive-sum-of-products (ESOP) form, then penalize that ESOP expression in the cost Hamiltonian. Using maximum independent set (MIS) on graphs with 3 to 20 vertices, the authors claim this encoding beats the standard quadratic penalty approach, with average approximation ratio increases of up to 30.3% and better performance on roughly 64% of graphs at one QAOA layer. If true, the result would show that the way a constraint is algebraically encoded—not just the QAOA circuit depth—can substantially affect solution quality, and that ESOP formulations might offer a general strategy for constrained optimization on quantum hardware.

What carries the argument

The central object is the ESOP (exclusive-sum-of-products) Boolean expression, a sum modulo 2 of products of literals. The pipeline converts the MIS constraint OR into this form, then maps each product term to a Pauli-Z Hamiltonian via the representation table of [15], with the trick of multiplying every second variable by −1 to keep the Hamiltonian Hermitian. The key simplification is Eq. (2): when two product terms contain a variable in both its positive and negative literal form, their Hamiltonian sum collapses to a sum of individual terms, eliminating high-weight operators and producing a tractable penalty Hamiltonian.

What would settle it

Check the ESOP-derived constraint function on a single bitstring: for a=b=1, the identity a∨b = ab⊕b gives 1∨1 = 1 but ab⊕b = 1⊕1 = 0, so the translated Hamiltonian marks a disallowed pair as allowed. Concretely, on the P4 graph with bitstring (x1=1, x2=1, x3=1, x4=0), the paper's ESOP expression evaluates to 1 (via the term x1∧x3) even though the set {x1,x2,x3} is not independent, contradicting the claimed equivalence.

Watch

Extended reading notes

Core claim

The central claim is that rewriting the MIS independence constraints as an ESOP expression—rather than using the standard QUBO penalty x_i x_j—produces a QAOA cost Hamiltonian whose approximation ratio is higher on most tested instances. The authors derive the ESOP form by applying the identity a∨b = ab⊕b recursively to the OR of all edge constraints, then convert each product term to a Hamiltonian using the rules of [15], simplifying the sum using the observation that complementary literals (x_j and x̄_j) make cross-terms vanish. They report that, across 54 configurations of graph size and QAOA depth p, the ESOP encoding wins 51 times, with the largest gains at p=1 on mid-sized graphs (28.7

Load-bearing premise

The load-bearing premise is the Boolean identity a∨b = ab⊕b, used to rewrite every OR of edge constraints into an ESOP form; if that step is incorrect, the resulting penalty Hamiltonian no longer encodes the independence constraint.

Editorial extensions

If this is right

  • If the claim holds, constraint encoding becomes a tunable part of QAOA design: the same MIS problem can be solved with better approximation ratios simply by rewriting the constraints as ESOP before penalization.
  • The method avoids ancillary qubits and keeps the cost Hamiltonian low-weight on graphs where complementary literals appear, which could reduce circuit depth and improve optimization landscapes.
  • The approach is stated for any constrained optimization problem, not just MIS, suggesting a general pipeline where Boolean constraints are first converted to ESOP and then penalized.
  • The reported improvements at p=1 are especially relevant since shallow QAOA circuits are the most practical near-term target.
  • The method's success on graphs up to 20 vertices suggests it may scale, though the paper notes that characterizing which graphs benefit is still open.

Reading between the lines

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

  • The paper's reported gains are contingent on the correctness of the ESOP derivation, which uses the identity a∨b = ab⊕b; that identity is algebraically false (for a=b=1, it gives 0 instead of 1), so the derived constraint Hamiltonian may not actually enforce independence on some inputs.
  • If the conversion is corrected to a valid ESOP (e.g., using proper De Morgan negations), the pipeline could still be viable, but the specific percentage improvements as stated would likely change and should be re-evaluated.
  • The simplification in Eq. (2) is special to MIS-like constraints where complementary literals appear; for other constraints the cross-terms may not vanish, so the circuit-depth advantage may not generalize without additional ESOP minimization.
  • A natural testable extension is to run the same pipeline on MaxCut with cycle constraints or on knapsack-type constraints, where the ESOP form could either help or hurt depending on literal complementarity.
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

4 major / 4 minor

Summary. The paper proposes an 'ESOP pipeline' for encoding constraints in QAOA: rewrite each constraint as a Boolean expression, convert it to exclusive-sum-of-products (ESOP) form, penalize the resulting expression in the cost Hamiltonian, and apply standard QAOA. The method is tested on the Maximum Independent Set (MIS) problem for connected graphs with 3–20 vertices, with simulations over thousands of instances and comparisons to the standard QUBO penalty approach. The authors report that the ESOP-encoded constraints yield higher average approximation ratios in 51 of 54 graph-size/layer configurations, with improvements up to 30.3%.

Significance. If correct, the paper would offer a practically useful and broadly applicable encoding technique for constrained QAOA, supported by an unusually extensive numerical study and by publicly available code. The writing is generally clear and the experimental framework is transparent. However, the central derivation of the ESOP constraint Hamiltonian contains load-bearing algebraic errors: the Boolean identity in §2.3 is false, the recursive expansion in §3.2 drops required negations, and the sign of the constraint penalty is inverted relative to the stated objective. As a result, the Hamiltonian actually simulated is not the Hamiltonian of the ESOP-encoded independence constraint, and the reported approximation-ratio improvements cannot be attributed to the claimed pipeline. The empirical study does not rescue the paper, because it tests the wrong cost function.

major comments (4)
  1. [§2.3] The displayed identity 'a ∨ b = ab ⊕ b = a ⊕ b ⊕ ab' is false. The first equality fails: for a=1,b=0, ab⊕b = 0 but a∨b=1. The correct ESOP identity is a∨b = a⊕b⊕ab = (a∧¬b)⊕b. Since this identity is the foundation of the recursive expansion in §3.2, the error propagates directly into the derived constraint expression.
  2. [§3.2] The recursive expansion of the OR of edge terms drops the required negations. Although the prose states 'a ∨ b = (a ∧ ¬b) ⊕ b', the displayed lines use '(e0 ∧ e1) ⊕ e1', '(e1 ∧ e2) ⊕ e2', etc., with no negations. Consequently the final ESOP expression is not equivalent to the OR of the constraints. Example: for e1=1,e2=0,e3=0, the expression '(e1∧e2∧e3)⊕(e2∧e3)⊕e3' evaluates to 0, while e1∨e2∨e3 evaluates to 1. The P4 computation in §3.4 inherits this error; for instance (x2∧x4∧x4∧x1∧x1∧x3) simplifies to x1∧x2∧x3∧x4, not x1∧x2∧x4, so the first term c1 is missing x3.
  3. [§3.1–§3.2] The Boolean function used for the penalty has the wrong meaning. Equation (3) defines vMIS = ∨_{ij∈E}(xi∧xj), i.e., the indicator that at least one edge is violated. Section 3.2 then states 'vIS = ∨_{ij∈E} xi∧xj', which identifies the independent-set indicator with the violation indicator. The complement of the OR is the AND of negated edge products, not the OR. Consequently Hc in §3.3 is the Hamiltonian of the violation indicator, and HC = -HMAX - 2|V|Hc lowers the energy when an edge is violated, rewarding infeasible states rather than penalizing them.
  4. [§3.3] The Hamiltonian substitution and the use of Eq. (2) are not valid for the derived ESOP terms. The rule 'multiply every second variable in the term by -1' changes the Boolean function being represented, so Hci is not the Hamiltonian of the corresponding ESOP product ci. Moreover, Eq. (2) requires the products Hfi Hfj to vanish for the pairs of terms; the terms obtained in §3.4 do not satisfy this. For example, c1=x1∧x2∧x4 and c2=x1∧x3∧x4 can both be true (x1=x2=x3=x4=1), so Hc1Hc2≠0 and the simplification H_{c1⊕c2⊕c3}=ΣHci does not follow.
minor comments (4)
  1. [§3.1] The line 'The independence constraints ... can be written as |E|^_{ij∈E} xi ∧ xj' omits the required negation; it should be ∧_{ij∈E} ¬(xi∧xj) (or a product of negations), otherwise it is not the independence condition.
  2. [§3.2] The labels vMIS and vIS are used inconsistently: the first displayed expansion says 'yields vMIS' but the expression is the OR of the edge terms, which was previously denoted vMIS only after Eq. (3). This sign/notation confusion contributes to the error in the penalty sign.
  3. [§3.4 / Fig. 1] The caption of Fig. 1 lists maximum independent sets as {x1,x2}, {x2,x3}, and {x3,x4} without explaining that the path order is x2-x4-x1-x3; as written, the sets appear to contain adjacent vertices and are confusing. Please clarify the vertex ordering.
  4. [General] Minor typos and formatting issues: 'independet' in §3.1, 'De’Morgans' without apostrophe, and the blue/red color indication in Table 2 is lost in grayscale. These do not affect the technical content.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported improvements are empirical comparisons; the algebraic defects are correctness risks, not circular reductions.

full rationale

The paper's central claim is empirical: ESOP-encoded constraint Hamiltonians yield higher QAOA approximation ratios than standard QUBO penalization on a suite of graphs. The derivation chain in Sections 3.1–3.3 converts a Boolean constraint expression into a candidate cost Hamiltonian and then simulates QAOA; no parameter is fitted to the reported outcome and then renamed a prediction. The penalty coefficient 2|V| and the (−1)^{j+1} sign rule are introduced as construction choices, not derived from the data, so even if the choices are arbitrary or wrong, they do not make the evaluation circular. The paper does cite prior work, including Hadfield [15] for the Boolean-to-Hamiltonian dictionary and BHT-QAOA [2] for the sign rule, but these are external sources and the present authors' related-work citations (e.g., [17,18,45]) are not load-bearing for the ESOP construction or the numerical comparison. The serious algebraic errors noted in Section 3.2/3.3 (e.g., a∨b = ab⊕b, dropped negations) are correctness defects, not circular reductions: the target result is not assumed as an input. Therefore no circular step can be exhibited, and the appropriate score is 0.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central derivation relies on a false Boolean identity, an unjustified sign alternation in the Hamiltonian mapping, and an incorrectly stated condition for term cancellation. These are hand-chosen rules and assumptions that are not independently grounded, and they determine the form of the cost Hamiltonian that drives the reported numerical results.

free parameters (2)
  • Penalty coefficient 2|V| = 2|V|
    The penalty strength in HC is chosen as twice the number of vertices without any derivation of sufficiency; this affects whether the energy landscape correctly penalizes infeasible states.
  • Sign alternation factor (-1)^{j+1} in H_ci = alternating signs
    The rule 'multiply every second variable in the term ci by -1' is introduced ad hoc to make higher-order terms cancel in the XOR-to-sum conversion; it is not derived from the Boolean expression and changes the meaning of the constraint term.
assumptions (6)
  • standard math Boolean identity a \vee b = a \oplus b \oplus ab (used but misapplied as ab \oplus b)
    The standard identity is correct, but the paper states a \vee b = ab \oplus b, which is false, and then builds the recursive ESOP expansion on this error (Section 2.3).
  • domain assumption Hamiltonian mapping rules from Hadfield [15] for Boolean expressions
    The paper relies on the table from [15] to convert Boolean expressions to Pauli Hamiltonians, but then deviates from it with the sign alternation rule without justification (Section 3.3).
  • standard math H_{f\oplus g} = H_f + H_g - 2H_f H_g
    This identity is referenced from [15] and used in Section 2.3; it is correct, but the condition for when the cross term vanishes is misstated in Eq. (2).
  • ad hoc to paper Eq. (2): cancellation of higher-order terms when a variable appears in opposite literals
    The stated condition says the set of such vertices has size > 1, but the P4 example has only one such vertex, and the theorem is used to assert that the Hamiltonian of the XOR equals the sum of individual Hamiltonians. This condition is not satisfied by the example terms.
  • ad hoc to paper Penalty coefficient 2|V| is sufficient to enforce independence constraints
    No analysis shows that 2|V| is large enough for all tested graphs; the choice is made by hand and could affect the approximation ratio comparison.
  • domain assumption COBYLA optimization finds good QAOA angles
    The results depend on classical optimization of gamma and beta; the paper does not study whether COBYLA converges to global optima or local minima, which could influence the reported approximation ratios.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Exclusive-Sum-of-Products Pipeline for QAOA." pith.science (2026). https://pith.science/paper/IUKKFGMX

@misc{pith2026250821686,
  author       = {Pith},
  title        = {Pith review of: An Exclusive-Sum-of-Products Pipeline for QAOA},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IUKKFGMX}},
  note         = {Machine review of arXiv:2508.21686}
}
read the original abstract

The quantum approximate optimization algorithm is commonly used to solve combinatorial optimization problems. While unconstrained problems map naturally into the algorithm, incorporating constraints typically requires penalizing constraint violations in the objective function. In this work, we propose an alternative approach that encodes constraints as Boolean expressions in exclusive-sum-of-products (ESOP) form before penalization. We test this method on the maximum independent set problem using graphs with 3 to 20 vertices and find that ESOP constraint formulations achieve higher approximation ratios than standard constraint penalization methods, with percent increases of up to 30.3%. Furthermore, ESOP constraint formulations result in higher approximation ratios than standard QAOA penalization approaches after one layer of the algorithm on approximately 64% of the tested graphs.

Figures

Figures reproduced from arXiv: 2508.21686 by the authors.

Figure 1
Figure 1. Path on four vertices, P4. This has three unique maximum independent sets: {x1, x2}, {x2, x3}, and {x3, x4}. 2.3 SOPs and ESOPs Boolean expressions are commonly written in a SOP formulation [1, 5, 22], in which products of Boolean variables (ANDs) are added together (ORs). Any Boolean expression can be converted to an equivalent ESOP, which is a Boolean expression consisting solely of exclusive sums of products of l… view at source ↗
Figure 2
Figure 2. shows the counts when simulating QAOA with this cost Hamiltonian. Note that the optimal solutions of {x1, x2}, {x2, x3}, and {x3, x4} correspond to bitstrings 1100, 0110, and 0011 respectively [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Comparison of approximation ratios between regular QAOA and our [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparison of approximation ratios between regular QAOA and our [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 34 canonical work pages

  1. [1]

    sum of products of sums

    Abhyankar, S.: Minimal“sum of products of sums" expressions of boolean functions. IRE Transactions on Electronic Computers (4), 268–276 (2009)

    work page 2009

  2. [2]

    Entropy 26(10), 843 (2024)

    Al-Bayaty, A., Perkowski, M.: Bht-qaoa: The generalization of quantum approxi- mate optimization algorithm to solve arbitrary boolean problems as hamiltonians. Entropy 26(10), 843 (2024)

    work page 2024

  3. [3]

    https://doi.org/10.48550/ ARXIV.2407.12587, https://arxiv.org/abs/2407.12587

    Allcock, J., Santha, M., Yuan, P., Zhang, S.: On the dynamical lie algebras of quantum approximate optimization algorithms (2024). https://doi.org/10.48550/ ARXIV.2407.12587, https://arxiv.org/abs/2407.12587

    arXiv doi:10.48550/arxiv.2407.12587 2024

  4. [4]

    arXiv preprint arXiv:2503.10077 (2025)

    Angara, P.P., Lykov, D., Stege, U., Alexeev, Y., Müller, H.: The art of avoiding constraints: A penalty-free approach to constrained combinatorial optimization with qaoa. arXiv preprint arXiv:2503.10077 (2025)

    arXiv 2025

  5. [5]

    IEEE Transactions on Computers 100(11), 1028–1039 (1978) 14 M

    Arevalo, Bredeson: A method to simplify a boolean function into a near minimal sum-of-products for programmable logic arrays. IEEE Transactions on Computers 100(11), 1028–1039 (1978) 14 M. Brunet et al

    work page 1978

  6. [6]

    In: 2020 IEEE International Conference on Quantum Computing and Engineering (QCE)

    Bärtschi,A.,Eidenbenz,S.:Grovermixersforqaoa:Shiftingcomplexityfrommixer design to state preparation. In: 2020 IEEE International Conference on Quantum Computing and Engineering (QCE). pp. 72–82. IEEE (2020)

    work page 2020

  7. [7]

    In: 2022 IEEE International Conference on Quantum Computing and Engineering (QCE)

    Bartschi, A., Eidenbenz, S.: Short-depth circuits for dicke state preparation. In: 2022 IEEE International Conference on Quantum Computing and Engineering (QCE). p. 87–96. IEEE (Sep 2022). https://doi.org/10.1109/qce53715.2022.00027, http://dx.doi.org/10.1109/QCE53715.2022.00027

    arXiv 2022

  8. [8]

    Physical Review A110(5), 052435 (2024)

    Brady, L.T., Hadfield, S.: Iterative quantum algorithms for maximum independent set. Physical Review A110(5), 052435 (2024)

    work page 2024

Show all 48 references
  1. [9]

    https://github.com/ mxttbrunet/Quantum-Walk-Project

    Brunet, M., Shah, S., Atallah, M.: Quantum-walk-project. https://github.com/ mxttbrunet/Quantum-Walk-Project

  2. [10]

    arXiv preprint arXiv:2504.08663 (2025)

    Bucher, D., Stein, J., Feld, S., Linnhoff-Popien, C.: If-qaoa: A penalty-free approach to accelerating constrained quantum optimization. arXiv preprint arXiv:2504.08663 (2025)

    2025

  3. [11]

    Science376(6598), 1209–1215 (2022)

    Ebadi,S.,Keesling,A.,Cain,M.,Wang,T.T.,Levine,H.,Bluvstein,D.,Semeghini, G., Omran, A., Liu, J.G., Samajdar, R., et al.: Quantum optimization of maximum independent set using rydberg atom arrays. Science376(6598), 1209–1215 (2022)

    2022

  4. [12]

    https://doi.org/10.48550/ARXIV.1411.4028, https://arxiv.org/abs/ 1411.4028

    Farhi, E., Goldstone, J., Gutmann, S.: A quantum approximate optimization algo- rithm (2014). https://doi.org/10.48550/ARXIV.1411.4028, https://arxiv.org/abs/ 1411.4028

    2014 doi arXiv

  5. [13]

    In: 2021 IEEE International Conference on Quantum Computing and Engineering (QCE)

    Golden, J., Bärtschi, A., O’Malley, D., Eidenbenz, S.: Threshold-based quantum optimization. In: 2021 IEEE International Conference on Quantum Computing and Engineering (QCE). pp. 137–147. IEEE (2021)

    2021

  6. [14]

    arXiv preprint arXiv:2409.18829 (2024)

    Goldstein-Gelb, B., Lotshaw, P.C.: Convergence guarantee for linearly-constrained combinatorial optimization with a quantum alternating operator ansatz. arXiv preprint arXiv:2409.18829 (2024)

    2024 arXiv

  7. [15]

    ACM Transactions on Quantum Computing2(4), 1–21 (2021)

    Hadfield, S.: On the representation of boolean and real functions as hamiltonians for quantum computing. ACM Transactions on Quantum Computing2(4), 1–21 (2021)

    2021

  8. [16]

    Algorithms12(2), 34 (2019)

    Hadfield, S., Wang, Z., O’gorman, B., Rieffel, E.G., Venturelli, D., Biswas, R.: From the quantum approximate optimization algorithm to a quantum alternating operator ansatz. Algorithms12(2), 34 (2019)

    2019

  9. [17]

    Scientific Reports12(1), 1–10 (2022)

    Herrman, R., Lotshaw, P.C., Ostrowski, J., Humble, T.S., Siopsis, G.: Multi- angle quantum approximate optimization algorithm. Scientific Reports12(1), 1–10 (2022)

    2022

  10. [18]

    Algorithms 14(10), 294 (2021)

    Herrman, R., Treffert, L., Ostrowski, J., Lotshaw, P.C., Humble, T.S., Siopsis, G.: Globally optimizing qaoa circuit depth for constrained optimization problems. Algorithms 14(10), 294 (2021)

    2021

  11. [19]

    Kazi, S., Larocca, M., Farinati, M., Coles, P.J., Cerezo, M., Zeier, R.: Analyzing the quantum approximate optimization algorithm: ansätze, symmetries, and lie algebras (2024), https://arxiv.org/abs/2410.05187

    2024

  12. [20]

    Nature Reviews Physics7(4), 174–189 (Mar 2025)

    Larocca,M.,Thanasilp,S.,Wang,S.,Sharma,K.,Biamonte,J.,Coles,P.J.,Cincio, L., McClean, J.R., Holmes, Z., Cerezo, M.: Barren plateaus in variational quantum computing. Nature Reviews Physics7(4), 174–189 (Mar 2025). https://doi.org/10. 1038/s42254-025-00813-9, http://dx.doi.org/...

    2025 doi

  13. [21]

    Frontiers in physics2, 74887 (2014)

    Lucas, A.: Ising formulations of many np problems. Frontiers in physics2, 74887 (2014)

    2014

  14. [22]

    IEEE transac- tions on computers48(3), 296–310 (2002)

    Luccio, F., Pagli, L.: On a new boolean function with applications. IEEE transac- tions on computers48(3), 296–310 (2002)

    2002

  15. [23]

    Quantum Information Processing18(3), 61 (2019) An Exclusive-Sum-of-Products Pipeline for QAOA 15

    Marsh, S., Wang, J.: A quantum walk-assisted approximate algorithm for bounded np optimisation problems. Quantum Information Processing18(3), 61 (2019) An Exclusive-Sum-of-Products Pipeline for QAOA 15

    2019

  16. [24]

    McKay,B.:Graphs[dataset],foundathttp://users.cecs.anu.edu.au/bdm/data/graphs.html

  17. [25]

    In: International Conference on Reversible Computation

    Meuli, G., Schmitt, B., Ehlers, R., Riener, H., De Micheli, G.: Evaluating esop optimization methods in quantum compilation flows. In: International Conference on Reversible Computation. pp. 191–206. Springer (2019)

    2019

  18. [26]

    arXiv preprint arXiv:2405.09169 (2024)

    Montanez-Barrera, J., Michielsen, K.: Towards a universal qaoa protocol: Evi- dence of quantum advantage in solving combinatorial optimization problems. arXiv preprint arXiv:2405.09169 (2024)

    2024 arXiv

  19. [27]

    Quantum Information Processing 24(5), 129 (2025)

    Montanez-Barrera, J., Willsch, D., Michielsen, K.: Transfer learning of optimal qaoa parameters in combinatorial optimization. Quantum Information Processing 24(5), 129 (2025)

    2025

  20. [28]

    Quantum Sci- ence and Technology7(4), 045036 (sep 2022)

    Ozaeta, A., van Dam, W., McMahon, P.L.: Expectation values from the single-layer quantum approximate optimization algorithm on ising problems. Quantum Sci- ence and Technology7(4), 045036 (sep 2022). https://doi.org/10.1088/2058-9565/ ac9013, https://doi.org/10.1088%2F2058-956...

    2022 doi

  21. [29]

    Jour- nal of Circuits, Systems, and Computers23(01), 1450015 (2014)

    Papakonstantinou, G.: A parallel algorithm for minimizing esop expressions. Jour- nal of Circuits, Systems, and Computers23(01), 1450015 (2014)

    2014

  22. [30]

    In: Proc

    Perkowski, M., Ross, T., Gadd, D., Goldman, J.A., Song, N.: Application of esop minimization in machine learning and knowledge discovery. In: Proc. Reed Muller. vol. 95 (1995)

    1995

  23. [31]

    arXiv preprint arXiv:1808.10816 (2018)

    Pichler, H., Wang, S.T., Zhou, L., Choi, S., Lukin, M.D.: Quantum optimiza- tion for maximum independent set using rydberg atom arrays. arXiv preprint arXiv:1808.10816 (2018)

    2018 arXiv

  24. [32]

    Quantum Information Processing 24(2), 60 (2025)

    Ponce, M., Herrman, R., Lotshaw, P.C., Powers, S., Siopsis, G., Humble, T., Os- trowski,J.:Graphdecompositiontechniquesforsolvingcombinatorialoptimization problems with variational quantum algorithms. Quantum Information Processing 24(2), 60 (2025)

    2025

  25. [33]

    In: Advances in optimization and nu- merical analysis, pp

    Powell, M.J.: A direct search optimization method that models the objective and constraint functions by linear interpolation. In: Advances in optimization and nu- merical analysis, pp. 51–67. Springer (1994)

    1994

  26. [34]

    Cambridge University Press, second edn

    Press, W.H., Flannery, B.P., Teukolsky, S.A.: Numerical Recipes in Fortran 77: The Art of Scientific Computing. Cambridge University Press, second edn. (1993), https://people.sc.fsu.edu/ inavon/5420a/DFP.pdf

    1993

  27. [35]

    https://quantum.cloud.ibm.com/docs/en/api/ qiskit/release-notes/0.16

    Qiskit: Qiskit 0.16 release notes. https://quantum.cloud.ibm.com/docs/en/api/ qiskit/release-notes/0.16

  28. [36]

    https://quantum.cloud.ibm.com/docs/en/api/ qiskit/release-notes/1.3

    Qiskit: Qiskit sdk 1.3 release notes. https://quantum.cloud.ibm.com/docs/en/api/ qiskit/release-notes/1.3

  29. [37]

    In: 2010 40th IEEE International Symposium on Multiple-Valued Logic

    Sanaee, Y., Dueck, G.W.: Esop-based toffoli network generation with transforma- tions. In: 2010 40th IEEE International Symposium on Multiple-Valued Logic. pp. 276–281. IEEE (2010)

    2010

  30. [38]

    In: 2019 IEEE 49th International Symposium on Multiple-Valued Logic (ISMVL)

    Schmitt, B., Soeken, M., De Micheli, G., Mishchenko, A.: Scaling-up esop syn- thesis for quantum compilation. In: 2019 IEEE 49th International Symposium on Multiple-Valued Logic (ISMVL). pp. 13–18. IEEE (2019)

    2019

  31. [39]

    Science Advances10(22) (2024)

    Shaydulin, R., Li, C., Chakrabarti, S., DeCross, M., Herman, D., Kumar, N., Lar- son, J., Lykov, D., Minssen, P., Sun, Y., et al.: Evidence of scaling advantage for the quantum approximate optimization algorithm on a classically intractable problem. Science Advances10(22) (2024)

    2024

  32. [40]

    In: Proceedings of the 14th ACM Great Lakes sym- posium on VLSI

    Stergiou, S., Daskalakis, K., Papakonstantinou, G.: A fast and efficient heuristic esop minimization algorithm. In: Proceedings of the 14th ACM Great Lakes sym- posium on VLSI. pp. 78–81 (2004) 16 M. Brunet et al

    2004

  33. [41]

    arXiv preprint arXiv:2010.14021 (2020)

    Tate, R., Farhadi, M., Herold, C., Mohler, G., Gupta, S.: Bridging classi- cal and quantum with sdp initialized warm-starts for qaoa. arXiv preprint arXiv:2010.14021 (2020)

    2010 arXiv

  34. [42]

    Quantum7, 1121 (2023)

    Tate, R., Moondra, J., Gard, B., Mohler, G., Gupta, S.: Warm-started qaoa with custom mixers provably converges and computationally beats goemans- williamson’s max-cut at low circuit depths. Quantum7, 1121 (2023)

    2023

  35. [43]

    Quantum Science and Technology 9(2), 025010 (2024)

    Vijendran,V.,Das,A.,Koh,D.E.,Assad,S.M.,Lam,P.K.:Anexpressiveansatzfor low-depth quantum approximate optimisation. Quantum Science and Technology 9(2), 025010 (2024)

    2024

  36. [44]

    Physical Review A 101(1), 012320 (2020)

    Wang, Z., Rubin, N.C., Dominy, J.M., Rieffel, E.G.: Xy mixers: Analytical and numerical results for the quantum alternating operator ansatz. Physical Review A 101(1), 012320 (2020)

    2020

  37. [45]

    arXiv preprint arXiv:2508.02590 (2025)

    Wilkie, A., DeLise, A., Del Real, A., Herrman, R., Ostrowski, J.: Learning feasible quantum states for quadratic constrained binary optimization problems. arXiv preprint arXiv:2508.02590 (2025)

    2025 arXiv

  38. [46]

    Physical Review A 110(2), 022441 (2024)

    Wilkie, A., Gaidai, I., Ostrowski, J., Herrman, R.: Quantum approximate opti- mization algorithm with random and subgraph phase operators. Physical Review A 110(2), 022441 (2024)

    2024

  39. [47]

    arXiv preprint arXiv:2504.21135 (2025)

    Xu, H., Liu, X., Pothen, A., Safro, I.: Qaoa parameter transferability for maximum independent set using graph attention networks. arXiv preprint arXiv:2504.21135 (2025)

    2025 arXiv

  40. [48]

    Physical Review Applied 19(2), 024027 (2023)

    Zhou, Z., Du, Y., Tian, X., Tao, D.: Qaoa-in-qaoa: solving large-scale maxcut problems on small quantum machines. Physical Review Applied 19(2), 024027 (2023)

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.