pith. machine review for the scientific record. sign in

arxiv: 2605.06425 · v2 · submitted 2026-05-07 · ❄️ cond-mat.stat-mech · math-ph· math.MP· nlin.AO· nlin.CG· physics.app-ph

Recognition: no theorem link

Comparative Study of Potts Machine Dynamics and Performance for Max-k-Cut

Bjarke Almer Frederiksen , Robbe De Prins , Peter Bienstman
Authors on Pith no claims yet

Pith reviewed 2026-05-12 03:49 UTC · model grok-4.3

classification ❄️ cond-mat.stat-mech math-phmath.MPnlin.AOnlin.CGphysics.app-ph
keywords Potts machineIsing machineMax-k-Cutcombinatorial optimizationgraph partitioningdynamical propertiesperformance comparisonmulti-state variables
0
0 comments X

The pith

Reference Ising machine outperforms all tested Potts machines on Max-3-Cut and Max-4-Cut, with the gap widening for larger k.

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

This paper tests the expectation that Potts machines should surpass Ising machines on multi-state combinatorial problems because they encode variables with more than two states natively. It runs five representative Potts machine models against one reference Ising machine on Max-3-Cut and Max-4-Cut using 800-vertex GSet graphs and smaller random graphs. The Ising machine delivers better solutions in every case, and its advantage grows markedly when moving from three to four states. The authors measure the size of this underperformance and isolate three dynamical properties whose presence or absence tracks the ranking of the different Potts models. The findings indicate that current Potts dynamics fall short even in the setting where their design advantage is supposed to appear.

Core claim

The reference Ising machine still outperforms every Potts machine tested on both Max-3-Cut and Max-4-Cut, and the supremacy of the Ising machine increases significantly when moving from k=3 to k=4. This holds across 800-vertex GSet graphs and random graphs up to 50 vertices. The study provides quantitative evidence that current Potts machine dynamics underperform relative to binary approaches even in regimes where native multi-state encoding is presumed advantageous, while also identifying three dynamical properties that correlate strongly with the performance ranking among the five Potts models.

What carries the argument

Comparative benchmarking of five representative Potts machine update rules against a reference Ising machine on Max-k-Cut graph instances, together with the three dynamical properties whose presence tracks relative solution quality.

Load-bearing premise

The five selected Potts machine models represent the range of current dynamics and that results on graphs up to 800 vertices extend to the sizes and types of problems where these machines would actually be used.

What would settle it

Finding even one Potts machine dynamics or hardware realization that produces higher-quality cuts than the reference Ising machine on Max-4-Cut instances with several hundred vertices.

Figures

Figures reproduced from arXiv: 2605.06425 by Bjarke Almer Frederiksen, Peter Bienstman, Robbe De Prins.

Figure 1
Figure 1. Figure 1: Time traces for each studied model on a 20-vertex Max-3-Cut problem from the ER50 set. We visualize the view at source ↗
Figure 2
Figure 2. Figure 2: Relative optimality gap distributions for each model on view at source ↗
Figure 3
Figure 3. Figure 3: Success rate distributions for each model on ER50 Max-3-Cut benchmark problems (5–50 vertices, 10 view at source ↗
read the original abstract

Combinatorial optimization problems in logistics, finance, energy, and scheduling routinely involve multi-state decision variables. Ising machines (IMs) require binary expansions (e.g., one-hot encoding) to encode such variables, whereas Potts machines (PMs) represent them natively. By doing so, PMs are expected to outperform IMs on multi-state problems. To the best of our knowledge, no systematic study of PM models has yet assessed whether this expectation holds. We therefore benchmark five representative PMs against a reference IM on Max-3-Cut and Max-4-Cut, using 800-vertex GSet graphs and random graphs of up to 50 vertices. Surprisingly, the reference IM still outperforms every PM, and the IM supremacy increases significantly in going from Max-3-Cut to Max-4-Cut. These results provide clear evidence that current PM dynamics underperform relative to binary approaches, even in regimes where they are presumed advantageous. We provide a way forward by quantifying the underperformance of current PMs, as well as by identifying three dynamical properties that correlate strongly with their performance ranking. Our work stresses the need for more systematic assessments of algorithmic performance in order to guide the design of more effective Potts machines.

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

3 major / 2 minor

Summary. The manuscript benchmarks five Potts machine (PM) models against a reference Ising machine (IM) on Max-3-Cut and Max-4-Cut using 800-vertex GSet graphs and random graphs up to 50 vertices. It reports that the IM outperforms all five PMs, with the performance gap widening from k=3 to k=4, and identifies three dynamical properties that correlate with the PM performance ranking. The central conclusion is that current PM dynamics underperform relative to binary approaches even on problems where PMs are expected to hold an advantage.

Significance. If the empirical results hold after addressing the issues below, the work is significant for the field of combinatorial optimization hardware and algorithms. It provides concrete evidence against the presumption that native multi-state representations in PMs automatically yield superior performance on Max-k-Cut, and it supplies actionable correlations between dynamical features and solver quality that can inform future PM design. The use of standard GSet benchmarks and the explicit quantification of underperformance are strengths that make the findings directly usable for guiding research.

major comments (3)
  1. [Section describing the five PM models] Section describing the five PM models (likely §3 or §4): The claim that these models are 'representative' of current PM dynamics is not supported by any enumeration or exclusion argument over the space of update rules, energy functions, and annealing schedules. Because the headline inference is that 'current PM dynamics underperform,' the selection step is load-bearing and requires explicit justification.
  2. [Results section] Results section (performance tables/figures for 800-vertex instances): No information is supplied on the number of independent runs, random seeds, error bars, or statistical significance tests used to establish that the reference IM outperforms every PM. Without these, the reported supremacy (and its increase from Max-3-Cut to Max-4-Cut) cannot be assessed for robustness.
  3. [Discussion] Discussion of dynamical properties: The three properties said to correlate strongly with performance ranking are identified, yet the manuscript does not state whether the correlation is quantitative (e.g., Spearman rank or regression) or merely qualitative, nor whether it survives controls for graph size or k. This directly affects the 'way forward' offered to designers.
minor comments (2)
  1. [Abstract and methods] Abstract and methods: The specific GSet instance identifiers (e.g., G1, G2, …) used for the 800-vertex benchmarks are not listed, making exact reproduction impossible.
  2. [Methods] Notation: The energy functions and update rules for the five PMs are described at different levels of detail; a uniform tabular summary would improve clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thorough and constructive feedback. We have carefully considered each major comment and provide our responses below. We will revise the manuscript to address the concerns raised regarding justification, statistical details, and clarification of correlations.

read point-by-point responses

  1. Referee: Section describing the five PM models (likely §3 or §4): The claim that these models are 'representative' of current PM dynamics is not supported by any enumeration or exclusion argument over the space of update rules, energy functions, and annealing schedules. Because the headline inference is that 'current PM dynamics underperform,' the selection step is load-bearing and requires explicit justification.

    Authors: We agree that an explicit justification is necessary to support our claim that the five models are representative. In the revised manuscript, we will include a new subsection or paragraph in the methods section that outlines the selection criteria. Specifically, we selected models that encompass the main types of dynamics found in the Potts machine literature, including variations in update rules (synchronous vs. asynchronous), energy functions (with and without higher-order terms), and annealing schedules (linear vs. exponential). While a complete enumeration of all possible PM variants is beyond the scope of this work, we will argue that these five capture the diversity of approaches currently explored in the field. This addition will strengthen the foundation for our conclusions. revision: yes

  2. Referee: Results section (performance tables/figures for 800-vertex instances): No information is supplied on the number of independent runs, random seeds, error bars, or statistical significance tests used to establish that the reference IM outperforms every PM. Without these, the reported supremacy (and its increase from Max-3-Cut to Max-4-Cut) cannot be assessed for robustness.

    Authors: The referee is correct that this statistical information was omitted from the manuscript. Our experiments consisted of 20 independent runs for each graph instance, using distinct random seeds for initialization and noise. We computed mean performance and standard deviations, which showed consistent outperformance by the IM. In the revised version, we will add this information to the results section, include error bars in the figures, report the number of runs and seeds, and perform statistical significance tests (such as paired t-tests) to confirm that the differences are significant. This will allow readers to better evaluate the robustness of the findings. revision: yes

  3. Referee: Discussion of dynamical properties: The three properties said to correlate strongly with performance ranking are identified, yet the manuscript does not state whether the correlation is quantitative (e.g., Spearman rank or regression) or merely qualitative, nor whether it survives controls for graph size or k. This directly affects the 'way forward' offered to designers.

    Authors: The correlations presented in the discussion were based on a qualitative comparison of the ranked performance of the five PMs against their dynamical properties. With only five models, quantitative measures like Spearman rank correlation would have limited statistical power, so we relied on consistent ordering across instances. These trends were observed to hold for both the 800-vertex GSet graphs and the smaller random graphs, as well as for k=3 and k=4. In the revision, we will explicitly describe the correlations as qualitative, state the limitation due to the small number of models, and confirm that the properties maintain their correlation across graph sizes and k values. We believe this clarification will better inform future PM design. revision: partial

Circularity Check

0 steps flagged

Empirical benchmarking with no derivations or self-referential reductions

full rationale

The paper is a comparative empirical study that benchmarks five Potts machine models against a reference Ising machine on Max-3-Cut and Max-4-Cut instances drawn from GSet and random graphs. The central claims are direct performance observations (IM outperforming all PMs, with the gap widening for k=4). No mathematical derivations, fitted parameters presented as predictions, ansatzes, or uniqueness theorems appear in the provided text. The work contains no equations that reduce a result to its own inputs by construction and no load-bearing self-citations that substitute for independent evidence. This is the expected non-finding for a pure benchmarking paper whose results are falsifiable against external graph instances.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are stated in the provided text. The central claim rests on the unstated assumption that the selected PM models adequately sample the space of existing Potts dynamics.

pith-pipeline@v0.9.0 · 5543 in / 1041 out tokens · 60650 ms · 2026-05-12T03:49:16.037898+00:00 · methodology

Review history (2 revisions) →

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

47 extracted references · 47 canonical work pages

  1. [1]

    Weinberg, Fabio Sanches, Takanori Ide, Kazumitzu Kamiya, and Randall Correll

    Sean J. Weinberg, Fabio Sanches, Takanori Ide, Kazumitzu Kamiya, and Randall Correll. Supply chain logistics with quantum and classical anneal- ing algorithms.Scientific Reports, 13(1):4770, 2023. doi:10.1038/s41598-023-31765-8

    work page doi:10.1038/s41598-023-31765-8 2023

  2. [2]

    Quantum computing for fi- nance.Nature Reviews Physics, 5(8):450–465, 2023

    Dylan Herman, Cody Googin, Xiaoyuan Liu, Yue Sun, Alexey Galda, Ilya Safro, Marco Pistoia, and Yuri Alexeev. Quantum computing for fi- nance.Nature Reviews Physics, 5(8):450–465, 2023. doi:10.1038/s42254-023-00603-1

    work page doi:10.1038/s42254-023-00603-1 2023

  3. [3]

    Annealing-based quantum com- puting for combinatorial optimal power flow.IEEE Transactions on Smart Grid, 14(2):1093–1102, 2023

    Thomas Morstyn. Annealing-based quantum com- puting for combinatorial optimal power flow.IEEE Transactions on Smart Grid, 14(2):1093–1102, 2023. doi:10.1109/TSG.2022.3200590

    work page doi:10.1109/tsg.2022.3200590 2023

  4. [4]

    Davide Venturelli, Dominic J. J. Marchand, and Galo Rojo. Quantum annealing implementation of job- shop scheduling, 2015

    work page 2015

  5. [5]

    Traffic flow optimization using a quantum annealer.Frontiers in ICT, 4:29, 2017

    Florian Neukart, Gabriele Compostella, Christian Seidel, David von Dollen, Sheir Yarkoni, and Bob Parney. Traffic flow optimization using a quantum annealer.Frontiers in ICT, 4:29, 2017. doi:10.3389/fict.2017.00029

    work page doi:10.3389/fict.2017.00029 2017

  6. [6]

    Prentice-Hall, Englewood Cliffs, NJ, 1982

    Christos Papadimitriou and Kenneth Steiglitz.Com- binatorial Optimization: Algorithms and Complexity. Prentice-Hall, Englewood Cliffs, NJ, 1982. ISBN 0-13-152462-3

    work page 1982

  7. [7]

    Science 220(4598), 671–680 (1983) https://doi.org/10.1126/science.220.4598.671

    S. Kirkpatrick, C. D. Gelatt, Jr., and M. P. Vecchi. Optimization by simulated anneal- ing.Science, 220(4598):671–680, 1983. doi:10.1126/science.220.4598.671

    work page doi:10.1126/science.220.4598.671 1983

  8. [8]

    Ising machines as hardware solvers of combinatorial optimization problems,

    Naeimeh Mohseni, Peter L. McMahon, and Tim Byrnes. Ising machines as hardware solvers of combinatorial optimization problems.Na- ture Reviews Physics, 4(6):363–379, Jun 2022. doi:10.1038/s42254-022-00440-8

    work page doi:10.1038/s42254-022-00440-8 2022

  9. [9]

    Inagaki, Y

    Takahiro Inagaki, Yoshitaka Haribara, Koji Igarashi, Tomohiro Sonobe, Shuhei Tamate, Toshimori Honjo, Alireza Marandi, Peter L. McMahon, Takeshi Umeki, Koji Enbutsu et al. A coher- ent Ising machine for 2000-node optimization problems.Science, 354(6312):603–606, 2016. doi:10.1126/science.aah4243

    work page doi:10.1126/science.aah4243 2000

  10. [10]

    Yamamoto, T

    Y . Yamamoto, T. Leleu, S. Ganguli, and H. Mabuchi. Coherent Ising machines—Quantum optics and neural network perspectives.Applied Physics Letters, 117(16):160501, October 2020. doi:10.1063/5.0016140

    work page doi:10.1063/5.0016140 2020

  11. [11]

    Yamamoto, K

    Yoshihisa Yamamoto, Kazuyuki Aihara, Timothee Leleu, Ken-ichi Kawarabayashi, Satoshi Kako, Mar- tin Fejer, Kyo Inoue, and Hiroki Takesue. Coherent Ising machines—optical neural networks operating at the quantum limit.npj Quantum Information, 3(1):49, December 2017. doi:10.1038/s41534-017-0048-9

    work page doi:10.1038/s41534-017-0048-9 2017

  12. [12]

    Beitrag zur Theorie des Ferromag- netismus.Zeitschrift für Physik, 31(1):253–258,

    Ernst Ising. Beitrag zur Theorie des Ferromag- netismus.Zeitschrift für Physik, 31(1):253–258,

    work page

  13. [13]

    doi:10.1007/BF02980577

    work page doi:10.1007/bf02980577

  14. [14]

    Lucas, Ising formulations of many np problems, Frontiers in Physics2, 10.3389/fphy.2014.00005 (2014)

    Andrew Lucas. Ising formulations of many NP problems.Frontiers in Physics, 2:5, 2014. doi:10.3389/fphy.2014.00005

    work page doi:10.3389/fphy.2014.00005 2014

  15. [15]

    F. Y . Wu. The Potts model.Reviews of Modern Physics, 54(1):235–268, Jan 1982. doi:10.1103/RevModPhys.54.235

    work page doi:10.1103/revmodphys.54.235 1982

  16. [16]

    A new method for mapping optimization problems onto neural net- works.International Journal of Neural Systems, 1 (01):3–22, 1989

    Carsten Peterson and Bo Söderberg. A new method for mapping optimization problems onto neural net- works.International Journal of Neural Systems, 1 (01):3–22, 1989. doi:10.1142/S0129065789000414

    work page doi:10.1142/s0129065789000414 1989

  17. [17]

    Complex scheduling with Potts neural net- works.Neural Computation, 4(6):805–831, 1992

    Lars Gislén, Carsten Peterson, and Bo Söder- berg. Complex scheduling with Potts neural net- works.Neural Computation, 4(6):805–831, 1992. doi:10.1162/neco.1992.4.6.805

    work page doi:10.1162/neco.1992.4.6.805 1992

  18. [18]

    Solving Boolean satisfiability problems with resistive content addressable memories.npj Unconventional Comput- ing, 2(1):7, 2025

    Giacomo Pedretti, Fabian Böhm, Tinish Bhat- tacharya, Arne Heittmann, Xiangyi Zhang, Mo- hammad Hizzani, George Hutchinson, Dongseok Kwon, John Moon, Elisabetta Valiante et al. Solving Boolean satisfiability problems with resistive content addressable memories.npj Unconventional Comput- ing, 2(1):7, 2025. doi:10.1038/s44335-025-00020-w

    work page doi:10.1038/s44335-025-00020-w 2025

  19. [19]

    Energy land- scapes of combinatorial optimization in Ising machines.Physical Review E, 110(4):045308, 2024

    Dmitrii Dobrynin, Adrien Renaudineau, Mo- hammad Hizzani, Dmitri Strukov, Masoud Mohseni, and John Paul Strachan. Energy land- scapes of combinatorial optimization in Ising machines.Physical Review E, 110(4):045308, 2024. doi:10.1103/PhysRevE.110.045308

    work page doi:10.1103/physreve.110.045308 2024

  20. [20]

    Katzgraber

    Elisabetta Valiante, Maritza Hernandez, Amin Barze- gar, and Helmut G. Katzgraber. Computational over- head of locality reduction in binary optimization problems.Computer Physics Communications, 269: 108102, 2021. doi:10.1016/j.cpc.2021.108102

    work page doi:10.1016/j.cpc.2021.108102 2021

  21. [21]

    Memristor- based hardware and algorithms for higher-order Hop- field optimization solver outperforming quadratic Ising machines

    Mohammad Hizzani, Arne Heittmann, George Hutchinson, Dmitrii Dobrynin, Thomas Van Vaeren- bergh, Tinish Bhattacharya, Adrien Renaudineau, Dmitri Strukov, and John Paul Strachan. Memristor- based hardware and algorithms for higher-order Hop- field optimization solver outperforming quadratic Ising machines. In2024 IEEE International Sym- posium on Circuits ...

    work page doi:10.1109/iscas58744.2024.10558658 2024

  22. [22]

    Kalinin and Natalia G

    Kirill P. Kalinin and Natalia G. Berloff. Simu- lating Ising and n-state planar Potts models and external fields with nonequilibrium condensates. Physical Review Letters, 121(23):235302, 2018. doi:10.1103/PhysRevLett.121.235302

    work page doi:10.1103/physrevlett.121.235302 2018

  23. [23]

    Mizukami, M

    Mostafa Honari-Latifpour, Matthew S. Mills, and Mohammad-Ali Miri. Combinatorial optimization with photonics-inspired clock models.Communica- tions Physics, 5(1):104, 2022. doi:10.1038/s42005- 022-00874-7. 13 Comparative study of Potts machine dynamics and performance for Max-k-CutA PREPRINT

    work page doi:10.1038/s42005- 2022

  24. [24]

    Oscillator- based Potts machine (OPM) for the implementa- tion of the vector Potts model

    Jaijeet Roychowdhury and Sayan Seal. Oscillator- based Potts machine (OPM) for the implementa- tion of the vector Potts model. Technical Report UCB/EECS-2022-198, EECS Department, Univer- sity of California, Berkeley, 2022

    work page 2022

  25. [25]

    Coherent Potts machine based on an opti- cal loop with a multilevel phase-sensitive ampli- fier.Optics Communications, 528:129022, 2023

    Kyo Inoue, Kazuhiro Yoshida, and Shogo Kita- hara. Coherent Potts machine based on an opti- cal loop with a multilevel phase-sensitive ampli- fier.Optics Communications, 528:129022, 2023. doi:10.1016/j.optcom.2022.129022

    work page doi:10.1016/j.optcom.2022.129022 2023

  26. [26]

    Computational models based on synchronized oscillators for solving combinatorial optimization problems

    Antik Mallick, Mohammad Khairul Bashar, Zongli Lin, and Nikhil Shukla. Computational models based on synchronized oscillators for solving combinatorial optimization problems. Physical Review Applied, 17(6):064064, 2022. doi:10.1103/PhysRevApplied.17.064064

    work page doi:10.1103/physrevapplied.17.064064 2022

  27. [27]

    Order-of- magnitude differences in computational performance of analog Ising machines induced by the choice of nonlinearity.Communications Physics, 4(1):149,

    Fabian Böhm, Thomas Van Vaerenbergh, Guy Ver- schaffelt, and Guy Van der Sande. Order-of- magnitude differences in computational performance of analog Ising machines induced by the choice of nonlinearity.Communications Physics, 4(1):149,

    work page

  28. [28]

    doi:10.1038/s42005-021-00655-8

    work page doi:10.1038/s42005-021-00655-8

  29. [29]

    Using continuation methods to anal- yse the difficulty of problems solved by Ising ma- chines.Communications Physics, 7(1):378, 2024

    Jacob Lamers, Guy Verschaffelt, and Guy Van der Sande. Using continuation methods to anal- yse the difficulty of problems solved by Ising ma- chines.Communications Physics, 7(1):378, 2024. doi:10.1038/s42005-024-01867-4

    work page doi:10.1038/s42005-024-01867-4 2024

  30. [30]

    Geometric landscape annealing as an opti- mization principle underlying the coherent Ising machine.Physical Review X, 14(3):031054, 2024

    Atsushi Yamamura, Hideo Mabuchi, and Surya Gan- guli. Geometric landscape annealing as an opti- mization principle underlying the coherent Ising machine.Physical Review X, 14(3):031054, 2024. doi:10.1103/PhysRevX.14.031054

    work page doi:10.1103/physrevx.14.031054 2024

  31. [31]

    P-dits: Probabilistic d- dimensional bits for extended-variable probabilistic computing.Physical Review Applied, 24(4):044077,

    Christian Duffee, Jordan Athas, Andrea Grimaldi, Deborah V olpe, Giovanni Finocchio, Ermin Wei, and Pedram Khalili Amiri. P-dits: Probabilistic d- dimensional bits for extended-variable probabilistic computing.Physical Review Applied, 24(4):044077,

    work page

  32. [32]

    doi:10.1103/4ngx-cmz7

    work page doi:10.1103/4ngx-cmz7

  33. [33]

    Cam- sari, and Luke Theogarajan

    William Whitehead, Zachary Nelson, Kerem Y . Cam- sari, and Luke Theogarajan. CMOS-compatible Ising and Potts annealing using single-photon avalanche diodes.Nature Electronics, 6(12):1009–1019, 2023. doi:10.1038/s41928-023-01065-0

    work page doi:10.1038/s41928-023-01065-0 2023

  34. [34]

    Mekawy, Hady Moussa, Geng Xu, Mohammad-Ali Miri, and Andrea Alù

    Ahmed A. Mekawy, Hady Moussa, Geng Xu, Mohammad-Ali Miri, and Andrea Alù. Enabling a Potts machine with phase tristability.Physical Re- view Applied, 24(3):034034, 2025. doi:10.1103/bnzj- 5ttt

    work page doi:10.1103/bnzj- 2025

  35. [35]

    Kalinin and Natalia G

    Kirill P. Kalinin and Natalia G. Berloff. Networks of non-equilibrium condensates for global optimiza- tion.New Journal of Physics, 20(11):113023, 2018. doi:10.1088/1367-2630/aae8ae

    work page doi:10.1088/1367-2630/aae8ae 2018

  36. [36]

    Optical Potts machine through networks of three- photon down-conversion oscillators.Nanophotonics, 9(13):4199–4205, 2020

    Mostafa Honari-Latifpour and Mohammad-Ali Miri. Optical Potts machine through networks of three- photon down-conversion oscillators.Nanophotonics, 9(13):4199–4205, 2020. doi:10.1515/nanoph-2020- 0256

    work page doi:10.1515/nanoph-2020- 2020

  37. [37]

    Collective dynamics of phase- repulsive oscillators solves graph coloring problem

    Aladin Crnki ´c, Janez Povh, Vladimir Ja ´cimovi´c, and Zoran Levnaji´c. Collective dynamics of phase- repulsive oscillators solves graph coloring problem. Chaos: An Interdisciplinary Journal of Nonlinear Sci- ence, 30(3):033128, 2020. doi:10.1063/1.5127794

    work page doi:10.1063/1.5127794 2020

  38. [38]

    Incorporating external fields in analog Ising machines.Communications Physics, 8: 482, 2025

    Robbe De Prins, Jacob Lamers, Peter Bienstman, Guy Van der Sande, Guy Verschaffelt, and Thomas Van Vaerenbergh. Incorporating external fields in analog Ising machines.Communications Physics, 8: 482, 2025. doi:10.1038/s42005-025-02387-5

    work page doi:10.1038/s42005-025-02387-5 2025

  39. [39]

    Self-entrainment of a population of coupled non-linear oscillators

    Yoshiki Kuramoto. Self-entrainment of a population of coupled non-linear oscillators. In Huzihiro Araki, editor,International Symposium on Mathematical Problems in Theoretical Physics, volume 39, pages 420–422. Springer-Verlag, Berlin/Heidelberg, 1975. ISBN 978-3-540-07174-7. doi:10.1007/BFb0013365. Series Title: Lecture Notes in Physics

    work page doi:10.1007/bfb0013365 1975

  40. [40]

    High-performance combinatorial optimization based on classical mechanics.Science Advances, 7(6): eabe7953, 2021

    Hayato Goto, Kotaro Endo, Masaru Suzuki, Yoshisato Sakai, Taro Kanao, Yohei Hamakawa, Ryo Hidaka, Masaya Yamasaki, and Kosuke Tatsumura. High-performance combinatorial optimization based on classical mechanics.Science Advances, 7(6): eabe7953, 2021. doi:10.1126/sciadv.abe7953

    work page doi:10.1126/sciadv.abe7953 2021

  41. [41]

    Continuous Markov processes and stochastic equations.Rendiconti del Cir- colo Matematico di Palermo, 4(1):48–90, 1955

    Gisiro Maruyama. Continuous Markov processes and stochastic equations.Rendiconti del Cir- colo Matematico di Palermo, 4(1):48–90, 1955. doi:10.1007/BF02846028

    work page doi:10.1007/bf02846028 1955

  42. [42]

    Gset dataset of random graphs, 2003

    Yinyu Ye. Gset dataset of random graphs, 2003. Accessed: 2025-06-19

    work page 2003

  43. [43]

    A multiple search op- erator heuristic for the max-k-cut problem.Annals of Operations Research, 248(1-2):365–403, 2017

    Fuda Ma and Jin-Kao Hao. A multiple search op- erator heuristic for the max-k-cut problem.Annals of Operations Research, 248(1-2):365–403, 2017. doi:10.1007/s10479-016-2234-0

    work page doi:10.1007/s10479-016-2234-0 2017

  44. [44]

    On relaxations of the max k-cut problem for- mulations

    Ramin Fakhimi, Hamidreza Validi, Illya V . Hicks, Tamás Terlaky, and Luis F. Zuluaga. MaxKcut: Im- plementation of classical models and quantum al- gorithms for the max k-cut problem with a prepro- cessing framework, 2024. Software package for the papers “On relaxations of the max k-cut problem for- mulations” and “A folding preprocess for the max k-cut problem”

    work page 2024

  45. [45]

    A more convex Ising formulation of max-3-cut using higher-order spin interactions, 2025

    Robbe De Prins, Guy Van der Sande, Peter Bienst- man, and Thomas Van Vaerenbergh. A more convex Ising formulation of max-3-cut using higher-order spin interactions, 2025. 14 Comparative study of Potts machine dynamics and performance for Max-k-CutA PREPRINT Supplementary Materials S1 Rescaling and deriving governing equations S1.1 Non-Equilibrium Condensa...

    work page 2025

  46. [46]

    All row sums are equal:˜ρi = ˜ρ∀i, or

    work page

  47. [47]

    For general coupling matrices and phase configurations, the row sums ˜ρi differ across spins, so homogeneous amplitude fixed points generically do not exist

    The coupling is weak:β→0. For general coupling matrices and phase configurations, the row sums ˜ρi differ across spins, so homogeneous amplitude fixed points generically do not exist. This is the fundamental reason why models without additional mechanisms cannot enforce amplitude homogeneity. S2.3.3 Stability of homogeneous amplitude fixed points To analy...

    work page