Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

This paper claims that reusing precomputed quantum-hardware embeddings cuts the total runtime of a hybrid quantum-classical Benders decomposition algorithm by about an order of magnitude, without harming solution quality, on transmission ne

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 09:17 UTC pith:73LCINYV

load-bearing objection Credible engineering contribution — first to reuse precomputed embeddings inside a Benders loop on D-Wave — but the headline speedup is computed only over successful runs, so the size of the improvement is not yet established. the 3 major comments →

arxiv 2601.14024 v3 pith:73LCINYV submitted 2026-01-20 quant-ph math.OC

Performance enhancing of hybrid quantum-classical Benders approach for MILP optimization

Sergio L\'opez-Ba\~nos , Elisabeth Lobe , Ontje L\"unsdorf , Oriol Ravent\'os This is my paper
classification quant-ph math.OC
keywords Benders decompositionquantum annealingQUBOembedding reusemixed-integer linear programmingtransmission network expansion planninghybrid quantum-classical algorithm
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to establish that the biggest practical bottleneck in a quantum-annealing-based Benders decomposition—recomputing the problem-to-hardware embedding at every iteration—can be removed by reusing a precomputed template embedding. On small transmission-network expansion planning instances, this swap reduces total wall-clock time by roughly tenfold while keeping solution quality within the same 5% band of a classical optimum. If correct, this makes hybrid quantum-classical Benders decomposition a more viable route for solving larger MILPs on current annealers. The authors also argue that total runtime then grows approximately linearly with the number of iterations, since solve time is dominated by fixed annealing parameters.

Core claim

The central claim is that two precomputed embedding strategies—one using a large fixed complete-graph template (FX) and one using the tightest complete graph matching the current QUBO size (TT)—reduce the total runtime of the BD-QA algorithm by approximately an order of magnitude compared with the default per-iteration embedding heuristic, without compromising solution quality. On a D-Wave-class quantum annealer tested with TNEP instances from 3 to 15 buses, the precomputed embedding strategies maintained or extended the range of problem sizes solved within the 5% acceptance threshold, and the runtime reduction was concentrated in the embedding preprocessing step, which is NP-hard. The paper

What carries the argument

Benders decomposition splits a mixed-integer linear program into a master problem holding all integer variables—reformulated as a quadratic unconstrained binary optimization (QUBO) problem—and a continuous linear subproblem solved classically. The load-bearing enhancement is embedding reuse: instead of computing a new minor embedding with a heuristic at every Benders iteration, the paper precomputes a complete-graph embedding for the annealer's hardware graph and uses either a fixed 160-vertex template (FX) or the tightest template matching the current QUBO size (TT). This moves the expensive, NP-hard embedding computation out of the iterative loop, leaving sampling time as the dominant cost

Load-bearing premise

The performance comparison averages only over runs that ended within 5% of the reference classical optimum, so failed runs are excluded from the time and iteration statistics, which can make the measured speedup look better than reality.

What would settle it

Run the three embedding strategies—default, fixed, and tightest—on a fresh batch of TNEP instances with varied demand distributions, and record total runtime and success rate over every run, including those that fail the 5% threshold. If the precomputed-embedding strategies do not show at least a several-fold total-time reduction or if their success rate falls below the default when failures are counted, the order-of-magnitude claim is refuted.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Per-iteration overhead of the quantum master problem becomes nearly independent of problem size, limited only by the fixed number of reads and annealing time.
  • The hybrid approach can be applied to larger MILPs on current quantum hardware, as long as the QUBO size stays within the precomputed template limit.
  • The algorithm is hardware-agnostic and can be combined with any QUBO solver, not just quantum annealers, so the embedding-reuse benefit may transfer to other iterative optimization loops.
  • Total runtime of the hybrid scheme scales roughly linearly with the number of Benders iterations, in contrast to the superlinear growth observed for classical exact solvers on the same test cases.

Where Pith is reading between the lines

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

  • The same embedding-reuse trick could accelerate other iterative quantum-classical schemes where the logical graph changes only by adding constraints, such as cut-generation loops or column-generation methods.
  • The benchmark averages performance only over runs that reached the 5% success threshold; if failed runs were included in the time and iteration statistics, the apparent order-of-magnitude speedup could shrink, so the claim needs a full-run comparison to be robust.
  • The fixed penalty weight, chosen by scaling with the optimal objective of a representative instance, might be made adaptive using dual information from each Benders cut, which could improve convergence without added qubits.
  • The paper's observed linear scaling suggests that further gains could come from dynamically tuning reads and annealing time per iteration, spending more samples only when the QUBO grows.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper presents a hybrid quantum-classical Benders' decomposition (BD-QA) for MILP, in which the integer master problem is reformulated as a QUBO and solved on a D-Wave quantum annealer, while the continuous subproblem is solved classically. The main technical contributions are: (i) two precomputed embedding strategies (FX and TT) intended to replace the default minorminer embedding in each Benders iteration, (ii) a conservative integer rounding of cut bounds to reduce slack variables, and (iii) a stopping rule based on a maximum QUBO size. The algorithm is benchmarked on transmission network expansion planning (TNEP) instances with 3–15 buses, using Gurobi as the reference. The central claim is that FX/TT embeddings reduce total runtime by about an order of magnitude relative to minorminer without degrading solution quality, as shown in Figures 3 and 7.

Significance. If the central claim holds, the paper makes a practically useful contribution: embedding reuse is a simple way to remove a major preprocessing bottleneck in quantum-annealer-based Benders decomposition, and the paper is, to my knowledge, the first to systematically study this effect. The implementation is made available in the quacla library, and the paper provides a detailed, reproducible experimental pipeline. The conservative cut-rounding and QUBO-size stopping criterion are also useful engineering ideas. However, the quantitative speedup claim is currently supported only by statistics conditioned on successful runs, and the penalty parameter is chosen with the help of classical optimal values, which weakens the strength of the benchmark conclusions.

major comments (3)
  1. [Section 3.5 and Figures 3, 6, 7] The headline speedup of FX/TT over MM is computed only from 'the first 10% successful runs' (Section 3.5), while all failed runs are excluded from time and iteration statistics. Since Figure 3 shows that MM has a lower success rate at larger sizes, its mean total time in Figure 7 is conditional on success and therefore biased in favor of FX/TT. The paper should report all-run statistics (e.g., expected total time including failed runs, or time-to-success with a fixed failure penalty) to support the order-of-magnitude claim. As written, the quantitative comparison does not reflect the real cost of using each embedding strategy.
  2. [Section 3.1, penalty weight P] The penalty weight P is set using the Gurobi optimal value of a representative instance as an initial guess, then the text notes that 'subsequent manual tuning indicated that a better penalty choice was possible, although this refined setting is not included'. This makes the reported results dependent on an unstated manual tuning process, and the claim that P is 'problem-independent' is not established. The paper should report the chosen P, the range explored, and the sensitivity of success rates and total time to P. Without this, it is unclear whether the FX/TT advantage is robust across reasonable penalty choices.
  3. [Section 5, convergence caveat] The paper correctly states that the heuristic MP solver does not guarantee convergence of Benders decomposition. This is not by itself a flaw, but it interacts with the success-based benchmarking: iterations that revisit the same integer solution due to rounding or penalty weakness may be excluded or counted differently across configurations. The manuscript would be strengthened by reporting how often the algorithm terminated by the QUBO-size limit versus by the gap criterion, separately for each embedding strategy, since this affects the meaning of the reported iteration counts.
minor comments (4)
  1. [Throughout] Several typographical issues: 'pyPSAPython' (Section 3.2), 'quaclalibrary' (Section 3.3), and inconsistent spacing in citations. Figure captions say 'Error bars represent the standard deviation across 10% successful runs' — this is ambiguous and should be rephrased.
  2. [Algorithm 1] The return statement 'return (x_i, y_i, z, z)' uses the same symbol for upper and lower bound; please clarify which bound is returned and why both are needed.
  3. [Figure 6/7 captions] The captions state 'Error bars represent the standard deviation across 10% successful runs'. Since only the first 10% of successful runs are selected, this is not a standard deviation of the full distribution; consider reporting the entire distribution or at least stating the selection rule in the caption.
  4. [Section 3.1, formula for α bounds] The definitions of α and α contain the symbols λ_j^− and λ_j^+; these should be defined explicitly (sum of negative and positive components, presumably of the vector λ_j) to avoid confusion.

Circularity Check

0 steps flagged

No derivation-level circularity: the embedding-speedup claim is an independent empirical benchmark, with only minor self-citation in the background.

full rationale

The paper's load-bearing result is empirical: for a fixed BD-QA pipeline, replacing iterative minorminer embeddings with FX/TT precomputed complete-graph embeddings reduces total runtime by about an order of magnitude (Figures 6 and 7) without degrading solution quality (Figure 2). This comparison is measured end-to-end on D-Wave hardware; no equation defines the speedup in terms of the precomputation time of the embeddings, and no fitted parameter is relabeled as a prediction. The penalty weight P is calibrated using a Gurobi optimal value on a representative instance and then fixed across configurations; it is a hyperparameter of the experiment, not the output being predicted, so this calibration is not circular. The precomputed-embedding technique is attributed to the authors' prior work [35], and NP-hardness of minor embedding to [29]; these are self-citations, but they supply a method and a motivation, not the measured speedup itself, and the cited results are externally checkable. The benchmarking decision to average only the first 10% of successful runs (relative error below 5% of Gurobi) is a genuine limitation, since it excludes failed runs from time statistics and may bias the speedup comparison, but that is a statistical/correctness concern rather than a definitional reduction of an output to an input. No specific circular step can be exhibited, so the score reflects only the presence of minor background self-citation.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claim rests on standard Benders/QUBO math plus two domain-specific choices: the feasibility-by-load-shedding property of TNEP and the conservative rounding of cuts. The empirical evaluation introduces several hand-set parameters, most importantly the QUBO penalty P, which is fitted against a classical Gurobi solution and affects all downstream success metrics. No new physical entities are postulated.

free parameters (4)
  • QUBO penalty weight P = not stated; scaled as P ∝ 1/α with α ≈ 3100 from a classical Gurobi solution of a representative instance
    Section 3.1: 'penalty parameter P was selected empirically to obtain a setting that performs robustly across most instances' and 'α was first replaced by the optimal value obtained classically using Gurobi on a representative instance.' This parameter directly affects solution quality and success rate.
  • Stopping gap ε = 5%
    Section 3.1: 'a stopping criterion terminates the procedure if the gap is less than or equal to 5%.' This is a user-chosen tolerance, not derived, and it defines what counts as a 'successful' run.
  • Maximum QUBO size q_max = 160
    Section 3.1: q_max is set to 160, 'close to the maximum number of vertices in a complete graph that can be embedded in the current hardware.' It is a hardware-dependent cutoff rather than a tuned parameter, but it affects the number of iterations and the reported success behavior.
  • Annealing parameters RUNS, READS, SWEEPS = RUNS=100, READS=100; SWEEPS=100 for SA
    Section 3.5: parameters are fixed for all runs; the paper says SWEEPS=100 'is found to be a good enough parameter' and the main configuration is chosen as 'a reasonable trade-off'. These are hand-picked and affect runtime and solution quality.
axioms (4)
  • standard math Benders decomposition converges for this MILP if the master problem is solved exactly (Geoffrion's theorem).
    Section 2.1: 'The convergence behaviour of the BD in this MILP formulation is guaranteed for deterministic solvers [31, Theorem 2.4].' The paper then relaxes this by using a heuristic annealer, which breaks the premise.
  • domain assumption All TNEP subproblems are feasible for every binary assignment because load-shedding variables exist.
    Section 2.4: 'all possible binary variable assignments are feasible... due to the load shedding variable. Therefore... the cuts generated by the BD are always optimality cuts.' This assumption restricts the algorithm to problems with this feasibility property.
  • ad hoc to paper A conservative integer rounding of cut bounds (η_j = floor(s_j)) and keeping unrounded λ does not exclude valid master-problem solutions.
    Section 3.1: 'the bounds are conservatively rounded to an integer in each iteration... such that valid solutions of the MP are not excluded.' This is an algorithmic design choice, plausible but not formally proven; the paper notes that rounding errors may cause redundant cuts.
  • ad hoc to paper The QUBO penalty reformulation with a single fixed penalty P, scaled with 1/α, yields feasible and near-optimal solutions of the original master problem.
    Section 3.1: the penalty choice is justified by the observation that the constraint term scales as α² while the objective scales as α, so P∝1/α. The paper explicitly says the parameter was selected empirically and that better settings were found but not used.

pith-pipeline@v1.3.0-alltime-deepseek · 22571 in / 7042 out tokens · 71431 ms · 2026-08-03T09:17:38.672663+00:00 · methodology

0 comments
read the original abstract

Mixed-integer linear programming problems are extensively used in industry for a wide range of optimization tasks. However, as they get larger, they present computational challenges for classical solvers within practical time limits. Quantum annealers can, in principle, accelerate the solution of problems formulated as quadratic unconstrained binary optimization instances, but their limited scale currently prevents achieving practical speedups. Quantum-classical algorithms have been proposed to take advantage of both paradigms and to allow current quantum computers to be used in larger problems. In this work, a hardware-agnostic Benders' decomposition algorithm and a series of enhancements with the goal of taking the most advantage of quantum computing are presented. The decomposition consists of a master problem with integer variables, which is reformulated as a quadratic unconstrained binary optimization problem and solved with a quantum annealer, and a linear subproblem solved by a classical computer. The enhancements consist, among others, of different embedding processes that substantially reduce the pre-processing time of the embedding computation without compromising solution quality, a conservative handling of cut constraints, and a stopping criterion that accounts for the limited size of current quantum computers and their heuristic nature. The proposed algorithm is benchmarked against classical approaches using a D-Wave quantum annealer for a scalable family of transmission network expansion planning problems.

Figures

Figures reproduced from arXiv: 2601.14024 by Elisabeth Lobe, Ontje L\"unsdorf, Oriol Ravent\'os, Sergio L\'opez-Ba\~nos.

Figure 1
Figure 1. Figure 1: Diagram of the Benders’ Decomposition algorithm. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Average objective values compared to the state [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 5
Figure 5. Figure 5: Average (last) QUBO size of each problem instance [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Average number of iterations required for conver [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Average solve time (in seconds, in log scale) as [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Average total time (in seconds, in log scale) mea [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Objective value and success rate as functions of problem size across different solver configurations. Each row corresponds to a different number of reads and repetitions: (top) 100 runs with 10 reads and (bottom) 10 runs with 1000 reads. The left column shows the average objective values compared to the state-of-the-art solution and the 5% margin (shaded area), while the right column shows the correspondin… view at source ↗
Figure 10
Figure 10. Figure 10: Iteration count and QUBO size as functions of problem size across different solver configurations. Each row corresponds to a different number of reads and repetitions: (top) 100 runs with 10 reads, (middle) 100 runs with 100 reads, and (bottom) 10 runs with 1000 reads. The left column shows the average number of iterations required for convergence, while the right column reports the last QUBO size of each… view at source ↗
Figure 11
Figure 11. Figure 11: Solve time and total time vs. problem size for various solvers and embedding strategies. Each row corresponds to a different number of reads and repetitions: (top) 100 runs with 10 reads and (bottom) 10 runs with 1000 reads. The left column shows the solve time (in seconds) as a function of the number of buses, and the right column shows the corresponding total time (in seconds), both plotted on logarithm… view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Hybrid quantum-classical end-to-end pipeline for solving MILPs: a vehicle routing case study

    quant-ph 2026-07 accept novelty 4.0

    Gate-based QAOA can replace annealing for cut selection in MCMS Benders on VRP, but at tested scales cut selection is not the classical bottleneck and QAOA is slower than Cbc.

Reference graph

Works this paper leans on

56 extracted references · 14 canonical work pages · cited by 1 Pith paper · 4 internal anchors

  1. [1]

    Quantum Computing in the NISQ Era and Beyond

    John Preskill. “Quantum Computing in the NISQ Era and Beyond”. In:Quantum2 (Aug. 2018), p. 79.doi: 10.22331/q-2018-08-06-79

    doi:10.22331/q-2018-08-06-79 2018

  2. [3]

    2014.doi: 10.48550/ARXIV.1411.4028

    Edward Farhi, Jeffrey Goldstone, and Sam Gutmann.A Quantum Approxi- mate Optimization Algorithm. 2014.doi: 10.48550/ARXIV.1411.4028

    Pith/arXiv arXiv doi:10.48550/arxiv.1411.4028 2014

  3. [4]

    A Variational Eigen- value Solver on a Photonic Quantum Pro- cessor

    Alberto Peruzzo et al. “A Variational Eigen- value Solver on a Photonic Quantum Pro- cessor”. In:Nature Communications5.1 (July 2014), p. 4213.doi: 10.1038/ncomms5213

    doi:10.1038/ncomms5213 2014

  4. [5]

    In:Proceedings of the Royal Society of London

    David Elieser Deutsch. In:Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences425.1868(Sept.1989),pp.73– 90.doi: 10.1098/rspa.1989.0099

    arXiv 1989

  5. [6]

    Quantum annealing with manufactured spins

    M. W. Johnson et al. “Quantum annealing with manufactured spins”. In:Nature473.7346 (May 2011), pp. 194–198.doi: 10.1038/nature10012

    doi:10.1038/nature10012 2011

  6. [7]

    2000.doi: 10.48550/ARXIV.QUANT-PH/0001106

    Edward Farhi et al.Quantum Computa- tion by Adiabatic Evolution. 2000.doi: 10.48550/ARXIV.QUANT-PH/0001106

    doi:10.48550/arxiv.quant-ph/0001106 2000

  7. [8]

    Dalzell et al.Quantum Algorithms: A Survey of Applications and End-to-End Complexities

    Alexander M. Dalzell et al.Quantum Algorithms: A Survey of Applications and End-to-End Complexities. 2023.doi: 10.48550/ARXIV.2310.03011

    Pith/arXiv arXiv doi:10.48550/arxiv.2310.03011 2023

  8. [9]

    doi: 10.48550/ARXIV.2312.02279

    Amira Abbas et al.Quantum Optimization: Po- tential, Challenges, and the Path Forward.2023. doi: 10.48550/ARXIV.2312.02279

    Pith/arXiv arXiv doi:10.48550/arxiv.2312.02279 2023

  9. [10]

    [software]https : / / github

    D-Wave Systems Inc.D-Wave Ocean SDK. [software]https : / / github . com / dwavesyste ms/dwave-ocean-sdk. 2024

    2024

  10. [11]

    Chal- lenges in the Decarbonization of the Energy Sector

    Elisa Papadis and George Tsatsaronis. “Chal- lenges in the Decarbonization of the Energy Sector”. In:Energy205 (2020), p. 118025.doi: 10.1016/j.energy.2020.118025

    arXiv 2020

  11. [12]

    Complexity profiles: A large- scale review of energy system models in terms of complexity

    Elias Ridha, Lars Nolting, and Aaron Praktiknjo. “Complexity profiles: A large- scale review of energy system models in terms of complexity”. In:Energy Strategy Reviews30 (July 2020), p. 100515.doi: 10.1016/j.esr.2020.100515

    arXiv 2020

  12. [13]

    Planning Low-Carbon Electricity Systems Under Uncertainty Consid- ering Operational Flexibility and Smart Grid Technologies

    Rodrigo Moreno et al. “Planning Low-Carbon Electricity Systems Under Uncertainty Consid- ering Operational Flexibility and Smart Grid Technologies”. In:Philosophical Transactions of the Royal Society A: Mathematical, Physi- cal and Engineering Sciences375.2100 (2017), p. 20160305.doi: 10.1098/rsta.2016.0305

    arXiv 2017

  13. [14]

    Recent Advances in Energy Storage Systems for Renewable Source Grid Integration: A Comprehensive Review

    Muhammed Y. Worku. “Recent Advances in Energy Storage Systems for Renewable Source Grid Integration: A Comprehensive Review”. In:Sustainability14.10 (2022), p. 5985.doi: 10.3390/su14105985

    work page doi:10.3390/su14105985 2022

  14. [15]

    Wiley, June 1988.doi: 10.1002/9781118627372

    George Nemhauser and Laurence Wolsey.In- teger and Combinatorial Optimization. Wiley, June 1988.doi: 10.1002/9781118627372

    doi:10.1002/9781118627372 1988

  15. [16]

    Complexity of Branch- and-Bound and Cutting Planes in Mixed- Integer Optimization — II

    Amitabh Basu et al. “Complexity of Branch- and-Bound and Cutting Planes in Mixed- Integer Optimization — II”. In:Combina- torica42.S1 (Oct. 2022), pp. 971–996.doi: 10.1007/s00493-022-4884-7

    work page doi:10.1007/s00493-022-4884-7 2022

  16. [17]

    Partitioning procedures for solv- ing mixed-variables programming problems

    J. F. Benders. “Partitioning procedures for solv- ing mixed-variables programming problems”. In:Numerische Mathematik4.1 (Dec. 1962), pp. 238–252.doi: 10.1007/bf01386316

    doi:10.1007/bf01386316 1962

  17. [18]

    The De- composition Algorithm for Linear Programs

    George B. Dantzig and Philip Wolfe. “The De- composition Algorithm for Linear Programs”. In:Econometrica29.4 (Oct. 1961), p. 767.doi: 10.2307/1911818

    work page doi:10.2307/1911818 1961

  18. [19]

    The Lagrangian Relax- ation Method for Solving Integer Programming Problems

    Marshall L. Fisher. “The Lagrangian Relax- ation Method for Solving Integer Programming Problems”. In:Management Science27.1 (Jan. 1981), pp. 1–18.doi: 10.1287/mnsc.27.1.1

    doi:10.1287/mnsc.27.1.1 1981

  19. [20]

    Optimization by Simulated Annealing

    S. Kirkpatrick, C. D. Gelatt, and M. P. Vec- chi. “Optimization by Simulated Annealing”. In:Science220.4598 (May 1983), pp. 671–680. doi: 10.1126/science.220.4598.671

    doi:10.1126/science.220.4598.671 1983

  20. [21]

    A New Polynomial-Time Al- gorithm for Linear Programming

    N. Karmarkar. “A New Polynomial-Time Al- gorithm for Linear Programming”. In:Com- binatorica4.4 (Dec. 1984), pp. 373–395.doi: 10.1007/BF02579150

    doi:10.1007/bf02579150 1984

  21. [22]

    Hy- brid Quantum Benders’ Decomposition for Mixed-Integer Linear Programming

    Zhongqi Zhao, Lei Fan, and Zhu Han. “Hy- brid Quantum Benders’ Decomposition for Mixed-Integer Linear Programming”. In:2022 IEEE Wireless Communications and Network- ing Conference (WCNC). Apr. 2022.doi: 10.1109/WCNC51071.2022.9771632

    arXiv 2022

  22. [23]

    A hybrid Quantum-Classical Algorithm for Mixed-Integer Optimization in Power Systems

    Petros Ellinas, Samuel Chevalier, and Spy- ros Chatzivasileiadis.A hybrid Quantum- Classical Algorithm for Mixed-Integer Op- timization in Power Systems. 2024.doi: 10.48550/ARXIV.2404.10693

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2404.10693 2024

  23. [24]

    Integrating quantum and classical computing for multi-energy system optimization using Benders decomposition

    Ludger Leenders et al. “Integrating quantum and classical computing for multi-energy system optimization using Benders decomposition”. In: (2023).doi: 10.48550/ARXIV.2309.16363

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2309.16363 2023

  24. [25]

    Leveraging Quantum Computing for Accelerated Classical Algorithms in Power Systems Optimization

    Rosemary Barrass, Harsha Nagarajan, and Carleton Coffrin.Leveraging Quantum Com- puting for Accelerated Classical Algorithms in Power Systems Optimization. 2025.doi: 10.48550/ARXIV.2503.19112

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.19112 2025

  25. [26]

    Mixed-integer linear programming solver using Benders decomposition assisted by a neutral-atom quantum processor

    M. Yassine Naghmouchi and Wesley da Silva Coelho. “Mixed-integer linear programming solver using Benders decomposition assisted by a neutral-atom quantum processor”. In:Physi- cal Review A110.1 (2024).doi: 10.1103/Phys- RevA.110.012434

    doi:10.1103/phys- 2024

  26. [27]

    Yassine Naghmouchi, and Wesley Coelho.Enhancing the Per- 16 formance of Quantum Neutral-Atom- Assisted Benders Decomposition

    Anna Joliot, M. Yassine Naghmouchi, and Wesley Coelho.Enhancing the Per- 16 formance of Quantum Neutral-Atom- Assisted Benders Decomposition. 2025.doi: 10.48550/ARXIV.2503.03518

    work page doi:10.48550/arxiv.2503.03518 2025

  27. [28]

    Hybrid Quantum- Classical Multi-Cut Benders Approach with a Power System Application

    Nikolaos G. Paterakis. “Hybrid Quantum- Classical Multi-Cut Benders Approach with a Power System Application”. In:Computational Chemical Engineering172 (2023), p. 108161. doi: 10.1016/j.compchemeng.2023.108161

    arXiv 2023

  28. [29]

    Minor Embedding in Broken Chimera and Derived Graphs Is NP-complete

    Elisabeth Lobe and Annette Lutz. “Minor Embedding in Broken Chimera and Derived Graphs Is NP-complete”. en. In:Theoretical Computer Science989 (2024), pp. 1–23.doi: 10.1016/j.tcs.2023.114369

    arXiv 2024

  29. [30]

    Antonio J Conejo et al.Decomposition tech- niques in mathematical programming. en. Berlin, Germany: Springer, Feb. 2010.doi: 10.1007/3-540-27686-6

    work page doi:10.1007/3-540-27686-6 2010

  30. [31]

    Generalized Benders decom- position

    A. M. Geoffrion. “Generalized Benders decom- position”. In:Journal of Optimization Theory and Applications10.4 (Oct. 1972), pp. 237–260. doi: 10.1007/BF00934810

    doi:10.1007/bf00934810 1972

  31. [32]

    Ising Formulations ofMany NP Problems

    AndrewLucas. “Ising Formulations ofMany NP Problems”. English. In:Frontiers in Physics2 (Feb. 12, 2014).doi: 10.3389/fphy.2014.00005

    arXiv 2014

  32. [33]

    Quantum Bridge Analytics I: a tutorial on formulating and using QUBO models

    Fred Glover, Gary Kochenberger, and Yu Du. “Quantum Bridge Analytics I: a tutorial on formulating and using QUBO models”. In: 4OR17.4 (Nov. 2019), pp. 335–371.doi: 10.1007/s10288-019-00424-y

    doi:10.1007/s10288-019-00424-y 2019

  33. [34]

    Minor-embedding in adiabatic quantum computation: II. Minor-universal graph design

    Vicky Choi. “Minor-embedding in adiabatic quantum computation: II. Minor-universal graph design”. en. In:Quantum Inf. Pro- cess.10.3 (June 2011), pp. 343–353.doi: 10.1007/s11128-010-0200-3

    doi:10.1007/s11128-010-0200-3 2011

  34. [35]

    Embedding of Complete Graphs in Broken Chimera Graphs

    Elisabeth Lobe, Lukas Schürmann, and Tobias Stollenwerk. “Embedding of Complete Graphs in Broken Chimera Graphs”. In:Quantum In- formation Processing20.7(2021),pp.1–27.doi: 10.1007/s11128-021-03168-z

    work page doi:10.1007/s11128-021-03168-z 2021

  35. [36]

    Minor-embedding in adiabatic quantum computation: I. The parameter setting problem

    Vicky Choi. “Minor-embedding in adiabatic quantum computation: I. The parameter setting problem”. In:Quantum Information Processing 7.5 (2008), pp. 193–209.doi: 10.1007/s11128- 008-0082-9

    doi:10.1007/s11128- 2008

  36. [37]

    Optimal Sufficient Requirements on the Embedded Ising Problem in Polynomial Time

    Elisabeth Lobe and Volker Kaibel. “Optimal Sufficient Requirements on the Embedded Ising Problem in Polynomial Time”. In:Quantum In- formation Processing22.8(2023),pp.1–44.doi: 10.1007/s11128-023-04058-2

    work page doi:10.1007/s11128-023-04058-2 2023

  37. [38]

    Combinatorial Problems in Programming Quantum Annealers

    Elisabeth Lobe. “Combinatorial Problems in Programming Quantum Annealers”. PhD the- sis. Magdeburg: Otto-von-Guericke-Universität Magdeburg, 2022.doi: 10.25673/89443

    work page doi:10.25673/89443 2022

  38. [39]

    D-Wave Documentation

    D-Wave Systems Inc.Errors and Error Correc- tion. D-Wave Documentation. [Accessed: 2025- 03-10]https://docs.dwavequantum.com/en/ latest/quantum_research/errors.html

    2025

  39. [40]

    Complexity of transmission network expansion planning: NP- hardness of connected networks and MINLP evaluation

    David Oertel and R. Ravi. “Complexity of transmission network expansion planning: NP- hardness of connected networks and MINLP evaluation”. In:Energy Systems5.1 (Aug. 2013), pp. 179–207.doi: 10.1007/s12667-013- 0091-3

    work page doi:10.1007/s12667-013- 2013

  40. [41]

    Optimal Transmission Switching

    E. B. Fisher, R. P. O’Neill, and M. C. Fer- ris. “Optimal Transmission Switching”. In: IEEE Transactions on Power Systems23.3 (Aug. 2008), pp. 1346–1355.doi: 10.1109/TP- WRS.2008.922256

    arXiv 2008

  41. [42]

    Transmission expansion planning using benders decomposition and local branching

    Keshava Dilwali et al. “Transmission expansion planning using benders decomposition and local branching”. In:2016 IEEE Power and Energy Conference at Illinois (PECI). IEEE, Feb. 2016. doi: 10.1109/peci.2016.7459265

    arXiv 2016

  42. [43]

    PyPSA: Python for Power System Analysis

    T. Brown, J. Hörsch, and D. Schlachtberger. “PyPSA: Python for Power System Analysis”. In:Journal of Open Research Software6.4 (1 2018).doi: 10.5334/jors.188

    doi:10.5334/jors.188 2018

  43. [44]

    Version v0.1.0

    Jonas Hörsch et al.PyPSA-Eur: An Open Opti- misation Model of the European Transmission System (Dataset). Version v0.1.0. Jan. 2020. doi: 10.5281/zenodo.3601882

    work page doi:10.5281/zenodo.3601882 2020

  44. [46]

    Extending the D-Wave with support for Higher Precision Coefficients

    John E. Dorband.Extending the D-Wave with support for Higher Precision Coefficients. 2018. doi: 10.48550/ARXIV.1807.05244

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.05244 2018

  45. [47]

    Quacla: Quantum Classical Algorithms

    Sergio López-Baños et al. “Quacla: Quantum Classical Algorithms”. On Demand. 2025

    2025

  46. [48]

    Quan- tum Optimization Applications With Quark and Quapps: Bridging the Gap Between Ap- plication and Hardware

    Lukas Windgätter and Elisabeth Lobe. “Quan- tum Optimization Applications With Quark and Quapps: Bridging the Gap Between Ap- plication and Hardware”. In:IEEE Soft- ware42.5 (Sept. 2025), pp. 34–42.doi: 10.1109/MS.2025.3564146

    arXiv 2025

  47. [49]

    Version 1.2

    Elisabeth Lobe and Lukas Windgätter.Quark - QUantum Application Reformulation Kernel. Version 1.2. DLR-SC, 2025.doi: 10.5281/ZEN- ODO.13944213

    doi:10.5281/zen- 2025

  48. [50]

    CVXPY: A Python-embedded modeling language for convex optimization

    Steven Diamond and Stephen Boyd. “CVXPY: A Python-embedded modeling language for convex optimization”. In:Journal of Machine Learning Research17.83 (2016), pp. 1–5.doi: 10.48550/arXiv.1603.00943

    Pith/arXiv arXiv doi:10.48550/arxiv.1603.00943 2016

  49. [51]

    2020.doi: 10.48550/ARXIV.2003.00133

    Kelly Boothby et al.Next-Generation Topol- ogy of D-Wave Quantum Processors. 2020.doi: 10.48550/ARXIV.2003.00133

    Pith/arXiv arXiv doi:10.48550/arxiv.2003.00133 2020

  50. [52]

    Nike Dattani, Szilard Szalay, and Nick Chancel- lor.Pegasus: The second connectivity graph for large-scale quantum annealing hardware. 2019. doi: 10.48550/ARXIV.1901.07636. [53]D-Wave Advantage TM System JUPSI. [Ac- cessed: 2025-08-06]https : / / www . fz - juel ich . de / en / ias / jsc / systems / quantum - 17 computing/juniq- facility/juniq/d- wave- ad...

    Pith/arXiv arXiv doi:10.48550/arxiv.1901.07636 2019

  51. [54]

    Macready, and Aidan Roy

    Jun Cai, William G. Macready, and Aidan Roy. A practical heuristic for finding graph minors. 2014.doi: 10.48550/ARXIV.1406.2741

    Pith/arXiv arXiv doi:10.48550/arxiv.1406.2741 2014

  52. [55]

    2025.doi: 10.48550/ARXIV.2504.13376

    Aitor Gomez-Tejedor, Eneko Osaba, and Esther Villar-Rodriguez.Addressing the Minor-Embedding Problem in Quantum Annealing and Evaluating State-of-the- Art Algorithm Performance. 2025.doi: 10.48550/ARXIV.2504.13376

    doi:10.48550/arxiv.2504.13376 2025

  53. [56]

    On the emerging poten- tial of quantum annealing hardware for combi- natorial optimization

    Byron Tasseff et al. “On the emerging poten- tial of quantum annealing hardware for combi- natorial optimization”. In:Journal of Heuris- tics30.5–6 (Aug. 2024), pp. 325–358.doi: 10.1007/s10732-024-09530-5

    doi:10.1007/s10732-024-09530-5 2024

  54. [57]

    [software]h ttps : / / github

    D-Wave Systems Inc.dwave-neal. [software]h ttps : / / github . com / dwavesystems / dwave - neal. 2024

    2024

  55. [58]

    Interior-Point Polynomial Algorithms in Convex Programming

    Yurii Nesterov and Arkadii Nemirovskii. Interior-Point Polynomial Algorithms in Convex Programming. Society for Industrial and Applied Mathematics, Jan. 1994.doi: 10.1137/1.9781611970791

    doi:10.1137/1.9781611970791 1994

  56. [59]

    Maximization of a linear function of variables subject to linear inequali- ties

    George B Dantzig. “Maximization of a linear function of variables subject to linear inequali- ties”. In:Activity analysis of production and al- location13 (1951), pp. 339–347. 18 A Cuts equivalence In this section, it is shown that the cuts generated by two equivalent formulations of the linear programming SP, are identical. First, consider the original S...

    1951