Pith. sign in

REVIEW 4 major objections 5 minor 47 references

Graph Coloring Approach to Solving Sudoku with Oscillatory Neural Networks

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read An added rule-violation repulsion term makes an oscillatory-network Sudoku solver near-perfect on 4x4 and accurate on 9x9 grids.

desk verdict Promising ONN Sudoku solver with a genuinely new mechanism, but the g-term stability claim is mathematically wrong and missing hyperparameters make the accuracy numbers hard to trust as reported. read the letter →

arxiv 2607.15814 v1 pith:D3CKVL5G submitted 2026-07-17 cs.LG

Filip Sabo , Aida Todri-Sanial This is my paper
classification cs.LG
keywords OscillatoryNeuralNetworksGraphColoringSudokuKuramotomodelMax-K-CutConstraintsatisfactionCombinatorialoptimizationPhasedynamics
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

The paper tries to establish that a Sudoku solver built from coupled phase oscillators can be made far more accurate by replacing an expensive Graph Coloring coupling with a cheaper Max-K-Cut term and adding a repulsion term that actively pushes the dynamics out of rule-violating configurations. On 4x4 puzzles the solver stays almost flawless across every tested fraction of unknown digits, and on 9x9 it reaches perfect accuracy up to 25% unknown digits and still solves about half of puzzles at 50% unknown. This matters because it suggests hard Sudoku-type constraints can be encoded directly into oscillator energy landscapes without explicit search or backtracking. If the mechanism works as claimed, it offers a recipe for adding hard constraints to oscillator-based combinatorial optimizers.

What carries the argument

The load-bearing construction is the modified phase ODE (Eq. 13). It contains: (1) a parameter-free repulsive coupling sin(K/2(phi_i - phi_j)), which costs less than the earlier f(Delta-phi) term and makes every K-th-root phase difference equally reachable; (2) a mask f_unknown that freezes known cells; and (3) the new rule-violation term K_G * sum_j W'_ij g(phi_i, phi_j), with g(phi_i, phi_j) = exp(-(1 - cos(phi_i - phi_j))/sigma), which fires whenever two cells sharing a row, column, or box settle on the same digit. This term converts the soft Max-K-Cut objective into a hard-constraint satisfier by kicking oscillators out of illegal phase configurations.

What would settle it

Take the 4x4 violating configuration shown in the paper (a 1 in the top-left cell while a known 1 already lies in the same column and box), initialize the Eq. 13 dynamics there with the reported parameter values, and integrate for longer than the benchmark's time horizon. If the network remains stuck at the violation or returns to it, the claim that the g-term kicks oscillators out of invalid states is false.

Watch

Extended reading notes

Core claim

The central claim is that the modified phase ODE (Eq. 13) solves Sudoku substantially more often than earlier HNN- and ONN-based solvers. The ODE freezes known cells, uses the cheaper sin(K/2(phi_i - phi_j)) coupling so all digit phases are equally reachable, and adds a repulsion term K_G * sum_j W'_ij g(phi_i, phi_j) with g(phi_i, phi_j) = exp(-(1 - cos(phi_i - phi_j))/sigma) that fires when two cells in the same row, column, or box settle on the same digit. Benchmarked on 1000 puzzles per condition, the solver achieves near-perfect accuracy for 4x4 Sudoku for unknown ratios from 0.125 to 0.875, and for 9x9 it reports 100% accuracy up to 25% unknown, above 80% at 37.5%, about 50% at 50%, an

Load-bearing premise

The central assumption is that the added repulsion term, with a finite peak width and a chosen strength, destabilizes every rule-violating phase configuration while leaving every valid solution stable; the paper gives no proof of this, only the claim.

Editorial extensions

If this is right

  • For 4x4 Sudoku, accuracy is nearly 100% across all tested unknown ratios, not just sparse puzzles.
  • For 9x9 Sudoku, the solver is perfect up to 25% unknown cells and remains above 80% accuracy at 37.5% unknown.
  • The solver breaks down for 9x9 puzzles with more than about 70% unknown cells, suggesting a hard limit of the present parameterization.
  • High accuracy is accompanied by order parameters near 1, so the order parameter can serve as a cheap monitor of whether the network has converged to a valid digit assignment.
  • Adding a constraint-violation repulsion term is a general recipe that may transfer to other constrained optimization problems representable as coloring.

Reading between the lines

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

  • We infer the repulsion term is a general 'same-label repulsion' that could be applied to other graph-coloring-based problems, such as register allocation or timetabling, by swapping the digit-distance metric for the relevant coloring distance.
  • We infer the sharp drop on 9x9 above 50% unknown may reflect insufficient integration time or an undertuned K_G rather than a fundamental failure; a longer time horizon or adaptive sigma might extend the operating range, and this is testable.
  • We infer that the near-perfect 4x4 behavior comes partly from small state spaces where the repulsion term has few violations to fight; the 9x9 results are a more honest stress test of the mechanism.
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 / 5 minor

Summary. This paper proposes an oscillatory neural network (ONN) solver for Sudoku formulated as a Max-K-Cut / graph-coloring problem. The contribution is a modified Kuramoto ODE (Eq. 13) that uses the cheaper interaction sin(K/2 (phi_i - phi_j)) instead of the earlier f-corrected term, freezes known cells via f_unknown, and adds a repulsion term K_G sum_j W'_ij g(phi_i, phi_j) intended to destabilize rule-violating configurations. The authors benchmark 4x4 and 9x9 puzzles at seven unknown ratios against an HNN solver and a prior ONN solver, reporting nearly flawless 4x4 accuracy and, for 9x9, perfect accuracy up to 25% unknown digits, roughly 80% at 37.5%, roughly 50% at 50%, and no solved puzzles above about 70% unknown. The numerical protocol uses a 250-puzzle training set for parameter selection and 1000 test puzzles, with order-parameter histograms reported for each setting. The central claim is that the additional g-term is what allows the proposed ONN to significantly outperform the two prior solvers in accuracy.

Significance. If the reported accuracy is robust, this is a useful empirical contribution to ONN-based combinatorial optimization: it shows that a simple additive obstruction term can improve constraint satisfaction in phase-based solvers for this benchmark, and it extends the existing Max-K-Cut/ONN mapping literature. The authors are honest about several limitations, including the four tunable parameters, the small training set, the degradation on 9x9 puzzles at high unknown ratios, and the lack of a convergence proof. However, the core mechanism is asserted rather than demonstrated: the 'steady state if and only if correctly filled' statement conflicts with the positivity of g for finite sigma, and missing hyperparameter, initialization, and integration details prevent independent verification. The paper provides no code, proofs, or other machine-checkable artifacts. With a fixed stability analysis and complete experimental disclosure, the contribution could become a solid empirical application paper; in its current form it is not fully substantiated.

major comments (4)
  1. [§III-A3 and §IV] The statement that the g-term in Eq. 13 'ensures a steady state if and only if all cells are correctly filled out' is not supported and is false for any finite sigma. In Eq. 16, g(phi_i, phi_j) = exp(-(1-cos(phi_i-phi_j))/sigma) > 0 for every distinct ideal-phase pair; for K=9 the minimal separation gives g = exp(-(1-cos(2*pi/9))/sigma) > 0. Hence at a correctly completed Sudoku, the contribution f_unknown K_G sum_j W'_ij g is strictly nonzero, so the ideal phases are not fixed points of Eq. 13. Accuracy therefore depends on the unstated integration horizon and stopping rule, and on sigma, K_G, K_S, and w_u. Please provide a fixed-point/stability analysis, or a controlled numerical demonstration that the drift is negligible over the chosen simulation time, and report the stopping criterion.
  2. [§III-B and §IV] The reported accuracies are not reproducible without the parameter-sweep details. The model has four tunable parameters (w_u, K_S, K_G, sigma), but the paper gives no final parameter values, no sweep ranges, no number of trials per parameter combination, and no ODE solver settings (method, dt, number of oscillatory cycles, stopping rule). The test set of 1000 puzzles is evaluated after a 'thorough sweep' on 250 training puzzles; without the selected parameter values and the actual training/test split, it is impossible to determine whether the accuracy is a generalized property or a selection artifact. Please tabulate the chosen parameters and the full integration protocol.
  3. [§III-A2, Eq. 11] The initialization of known cells is not specified. Eq. 11 sets f_unknown(phi_i)=0 for every known cell, so the known phases never evolve; they must be initialized at the corresponding digit phases from Eq. 4. The paper does not state this, nor whether known cells are reset after integration. If the initial phases were random, the solver could not use the clues; if they are fixed, this must be disclosed as part of the method, since it affects the 'solved' criterion and the comparison with baselines.
  4. [§IV, Fig. 3] The comparison with the conventional ONN solver appears to use accuracy values reported in [28] rather than a re-run on the same 1000-puzzle test set. Because [28] uses its own unknown-ratio grid (e.g., [0.1, 0.2, 0.3, 0.4, 0.5]) and possibly its own puzzle generator, the curves are not strictly comparable. The claim that the proposed solver 'significantly outperforms' should be demonstrated on identical puzzles with identical evaluation, or the discrepancy should be stated as a limitation. The HNN baseline is implemented here, but no implementation details (solver, iterations, parameter choices) are given either.
minor comments (5)
  1. [Eqs. 15–16] Eq. 15 idealizes g as 1 for |phi_i - phi_j| ≈ 2*pi*n and 0 otherwise, but Eq. 16 is a smooth positive function for any finite sigma. Specify the threshold used to classify a phase difference as a digit match and state the selected sigma.
  2. [§IV, Eq. 9] The order parameter kappa measures distance to ideal phases, so high accuracy mechanically implies high kappa. The statement that the g-term 'ensures' the observed accuracy-kappa relationship is an interpretation, not a test; a control experiment without the g-term would help distinguish mechanism from correlation.
  3. [Figs. 3–4] No error bars or confidence intervals are provided. With 1000 samples and seven unknown ratios, binomial confidence intervals would be a useful addition, especially for the 9x9 degradation region.
  4. [Eqs. 10 and 14] The set-builder notation 'i ∧ j ∈ row, column or box' and 'i ∧ j /∈ known' is ambiguous; define precisely which cells are connected and how known/unknown status is encoded. Also, 'the authors in [35]' should read 'the authors of [35]'.
  5. [General] No code or data-availability statement is included. Section III-B should also state the ODE integration method, time step, and total number of oscillatory cycles, since the results may be sensitive to stopping time.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central accuracy claim is an empirical benchmark against external baselines, not a derivation from the model's own fitted inputs.

full rationale

The paper's central claim is empirical: Eq. 13 defines a concrete ODE (a computationally cheaper Max-K-Cut term, known-cell freezing via f_unknown, and an additional violation-repulsion term K_G * sum W'_ij g(phi_i, phi_j)), and the reported result is measured accuracy on 1000 generated Sudoku puzzles after a parameter sweep on 250 training samples. No quantity presented as a prediction is fitted from the data it is said to predict. The g-term is an ansatz; its asserted property that it 'ensures a steady state if and only if all cells are correctly filled out' is an unproved design assumption, not an input to the benchmark. The order parameter kappa (Eq. 9) is defined as proximity to the ideal phases, so the observed kappa-accuracy association is partly definitional, but kappa is used only as a convergence diagnostic and is not the basis for the main accuracy comparison. Self-citations such as [9], [24], and [28] are background or baseline material and are not load-bearing; the prior ONN solver [28] is an external baseline whose reported accuracies are simply plotted. The paper's own stated limitations (small 250-sample training set, four tuned hyperparameters, speculation that the g-term may be mis-scaled or need more oscillatory cycles) are correctness/reproducibility concerns, not circularity. Even if the finite-sigma g-term perturbs valid solutions as the skeptical reading argues, that would undermine the theoretical 'iff' statement, not make the benchmark comparison circular.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical and depends on several tuned numerical parameters plus unproved dynamical assumptions. The fitted values are not reported, so the reader cannot see how much of the reported accuracy is a consequence of tuning. No new physical entities are introduced.

free parameters (5)
  • w_u
    Coupling strength between two unknown cells in the same row/column/box; constrained to be negative with |w_u| < 1; exact value not reported.
  • K_S
    Scaling factor of the subharmonic injection locking term sin(K phi_i) in Eq. 13; swept to force phases onto the K ideal values.
  • K_G
    Scaling factor of the rule-violation repulsion term involving g(phi_i, phi_j) in Eq. 13; tuned to destabilize invalid equal-phase configurations.
  • sigma
    Width of the peaks of g(phi_i, phi_j) in Eq. 16; controls how sharply equal phases are detected; exact value not reported.
  • ODE integration duration / oscillatory cycles
    The solver's convergence depends on how long Eq. 13 is integrated; the authors speculate more cycles are needed at high unknown ratios (Section V), but the value used in benchmarks is not stated.
assumptions (5)
  • domain assumption Kuramoto/ONN phase dynamics (Eqs. 2, 8) converge to phase-locked configurations corresponding to the roots of the right-hand side.
    Used throughout Section III to equate oscillator phases with Sudoku digits; standard in ONN literature, not proved here.
  • domain assumption Encoding Sudoku rules as negative all-to-all couplings within rows/columns/boxes (Eq. 10) with K=N colors yields a faithful Max-K-Cut formulation of the puzzle.
    Section III-A1 to III-A2; if this mapping is imperfect, solving the ODE solves a different problem.
  • ad hoc to paper The Gaussian-like g(phi_i, phi_j) in Eq. 16 is a sufficiently sharp and correctly scaled detector of forbidden equal phases, and the K_G term destabilizes exactly the invalid configurations.
    Section III-A3; this is the core new mechanism and is asserted rather than derived; its failure at high unknown ratios is visible in Fig. 3b.
  • ad hoc to paper Freezing known cells via f_unknown and assigning knowns a stronger coupling than unknowns is sufficient to propagate constraints from clues to empty cells.
    Section III-A2; no proof that the frozen knowns always anchor the correct attractor.
  • domain assumption The generated Sudoku puzzles have a unique solution, so 'accuracy' as exact match to the intended solution is well-defined.
    Benchmarks use PySudoku [39]; uniqueness of generated puzzles is not stated in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Graph Coloring Approach to Solving Sudoku with Oscillatory Neural Networks." pith.science (2026). https://pith.science/paper/D3CKVL5G

@misc{pith2026260715814,
  author       = {Pith},
  title        = {Pith review of: Graph Coloring Approach to Solving Sudoku with Oscillatory Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D3CKVL5G}},
  note         = {Machine review of arXiv:2607.15814}
}
abstract

Oscillatory Neural Networks (ONNs) present an attractive physics-based computing paradigm rooted in the dynamics of a network of typically fully coupled oscillators aiming to minimize an underlying energy function. In this paper, we propose an ONN-based solver for one well-known constrained combinatorial optimization problem, namely a Sudoku, by formulating the problem as a Graph Coloring problem. By modifying the already existing Graph Coloring solver to a computationally cheaper version and introducing an additional term ensuring the fulfillment of the Sudoku constraints, our solver is shown to significantly outperform the existing HNN- and ONN solvers in terms of accuracy. In particular, we are able to achieve nearly flawless accuracies on $4 \times 4$ as well as rather high accuracies on $9 \times 9$ Sudoku puzzles for different numbers of unknown digits.

Figures

Figures reproduced from arXiv: 2607.15814 by the authors.

Figure 1
Figure 1. Function f(∆ϕij ) from Eq. 7 (left) and term sin (∆ϕij + f(∆ϕij )) (right) for K = 4. Clearly, utilizing the model from Eq. 6 has multiple draw￾backs. First of all, the function in Eq. 7 has a tuneable parameter σ, which controls the width of the peaks. Second, computing the terms in Eq. 7 is computationally expensive, particularly when done so for multiple thousands of time steps and large problems. However, most i… view at source ↗
Figure 2
Figure 2. Example of a 4×4 Sudoku grid with a violation of the rules, where the thick numbers indicate the known values. Clearly, the digit 1 in the upper left hand side corner violates two rules simultaneously, as there is already a known digit 1 in the same column as well as box. However, despite breaking the rules, the ODE presented in Eq. 11 would produce a vanishing gradient, for the digits 1 and 1 are indeed 2πn K apart… view at source ↗
Figure 3
Figure 3. Accuracy as a function of unknown ratio of the HNN-Sudoku solver [36], the conventional ONN-Sudoku solver [28] and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Order parameters associated with the best accuracies achieved by our proposed ONN-Sudoku solver plotted as histograms [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 2 canonical work pages

  1. [28]

    Solving sudoku using oscillatory neural net- works,

    B. F. Haverkort, F. Sbravati, S. Porfir, and A. Todri- Sanial, “Solving sudoku using oscillatory neural net- works,”Neuromorphic Computing and Engineering, 2025

    2025

  2. [1]

    Yang and S

    X.-S. Yang and S. Koziel,Computational optimization and applications in engineering and industry. Springer Science & Business Media, 2011, vol. 359

    2011

  3. [2]

    An effective hybrid immune-hill climbing optimization approach for solving design and manufac- turing optimization problems in industry,

    A. R. Yıldız, “An effective hybrid immune-hill climbing optimization approach for solving design and manufac- turing optimization problems in industry,”Journal of Materials Processing Technology, vol. 209, no. 6, pp. 2773–2780, 2009

    2009

  4. [3]

    Fathi, M

    M. Fathi, M. Khakifirooz, and P. M. Pardalos,Optimiza- tion in large scale problems: Industry 4.0 and Society 5.0 applications. Springer, 2019, vol. 152

    2019

  5. [4]

    Bangert,Optimization for industrial problems

    P. Bangert,Optimization for industrial problems. Springer Science & Business Media, 2012

    2012

  6. [5]

    Reducibility among combinatorial prob- lems,

    R. M. Karp, “Reducibility among combinatorial prob- lems,” in50 Years of Integer Programming 1958-2008: from the Early Years to the State-of-the-Art. Springer, 2009, pp. 219–241

    1958

  7. [6]

    First draft of a report on the edvac,

    J. von Neumann, “First draft of a report on the edvac,” IEEE Annals of the History of Computing, vol. 15, no. 4, pp. 27–75, 1993

    1993

  8. [7]

    A survey of different approaches for overcoming the processor- memory bottleneck,

    D. Efnusheva, A. Cholakoska, and A. Tentov, “A survey of different approaches for overcoming the processor- memory bottleneck,”International Journal of Computer Science and Information Technology, vol. 9, no. 2, pp. 151–163, 2017

    2017

Show all 47 references
  1. [8]

    Deep learning’s carbon emissions problem,

    R. Toews, “Deep learning’s carbon emissions problem,”Forbes. [Online]. Available: https://www.forbes.com/sites/robtoews/2020/06/17/deep- learnings-climate-change-problem/

    2020

  2. [9]

    Computing with oscillators from theoretical underpin- nings to applications and demonstrators,

    A. Todri-Sanial, C. Delacour, M. Abernot, and F. Sabo, “Computing with oscillators from theoretical underpin- nings to applications and demonstrators,”Npj unconven- tional computing, vol. 1, no. 1, p. 14, 2024

    2024

  3. [10]

    The perceptron: a probabilistic model for information storage and organization in the brain

    F. Rosenblatt, “The perceptron: a probabilistic model for information storage and organization in the brain.” Psychological review, vol. 65, no. 6, p. 386, 1958

    1958

  4. [11]

    Yegnanarayana,Artificial neural networks

    B. Yegnanarayana,Artificial neural networks. PHI Learning Pvt. Ltd., 2009

    2009

  5. [12]

    Overview of artificial neural networks,

    J. Zou, Y . Han, and S.-S. So, “Overview of artificial neural networks,”Artificial neural networks: methods and applications, pp. 14–22, 2009

    2009

  6. [13]

    Simple model of spiking neurons,

    E. M. Izhikevich, “Simple model of spiking neurons,” IEEE Transactions on neural networks, vol. 14, no. 6, pp. 1569–1572, 2003

    2003

  7. [14]

    Oim: Oscillator-based ising machines for solving combinatorial optimisation problems,

    T. Wang and J. Roychowdhury, “Oim: Oscillator-based ising machines for solving combinatorial optimisation problems,” inUnconventional Computation and Natu- ral Computation: 18th International Conference, UCNC 2019, Tokyo, Japan, June 3–7, 2019, Proceedings 18. Springer, 2019, ...

    2019

  8. [15]

    Neural networks and physical systems with emergent collective computational abilities,

    J. Hopfield, “Neural networks and physical systems with emergent collective computational abilities,”Proceedings of the National Academy of Sciences of the United States of America, vol. 79, pp. 2554–8, 05 1982

    1982

  9. [16]

    Hopfield networks is all you need,

    H. Ramsaueret al., “Hopfield networks is all you need,” 2021

    2021

  10. [17]

    Beitrag zur theorie des ferro-und paramag- netismus,

    E. Ising, “Beitrag zur theorie des ferro-und paramag- netismus,” Ph.D. dissertation, Grefe & Tiedemann Ham- burg, Germany, 1924

    1924

  11. [18]

    Ising formulations of many np problems,

    A. Lucas, “Ising formulations of many np problems,” Frontiers in physics, vol. 2, p. 74887, 2014

    2014

  12. [19]

    Pattern recog- nition via synchronization in phase-locked loop neu- ral networks,

    F. C. Hoppensteadt and E. M. Izhikevich, “Pattern recog- nition via synchronization in phase-locked loop neu- ral networks,”IEEE Transactions on Neural Networks, vol. 11, no. 3, pp. 734–738, 2000

    2000

  13. [20]

    Overcoming quadratic hardware scaling for a fully connected digital oscillatory neural network,

    B. F. Haverkort and A. Todri-Sanial, “Overcoming quadratic hardware scaling for a fully connected digital oscillatory neural network,”Frontiers in Neuroscience, vol. V olume 19 - 2025, 2026. [Online]. Available: https://www.frontiersin.org/journals/neuroscience/articles/10.338...

    2025

  14. [21]

    Digital implementation of oscillatory neural network for image recognition applications,

    M. Abernot, T. Gil, M. Jim ´enez, J. N ´u˜nez, M. J. Avel- lido, B. Linares-Barranco, T. Gonos, T. Hardelin, and A. Todri-Sanial, “Digital implementation of oscillatory neural network for image recognition applications,”Fron- tiers in Neuroscience, vol. 15, p. 713054, 2021

    2021

  15. [22]

    Implementation of fitzhugh-nagumo neurons using nanoscale vo 2 devices,

    O. Maher, F. Horst, F. Sabo, D. Fancheng, V . Bra- gaglia, S. Karg, M. Sousa, A. Todri-Sanial, and B. J. Offrein, “Implementation of fitzhugh-nagumo neurons using nanoscale vo 2 devices,” IEEE, pp. 729–732, 2024

    2024

  16. [23]

    Learning at the speed of physics: Equilibrium propagation on oscillator ising machines,

    A. Gower, “Learning at the speed of physics: Equilibrium propagation on oscillator ising machines,” 2025. [Online]. Available: https://arxiv.org/abs/2510.12934

    2025

  17. [24]

    Classonn: Classification with oscillatory neural networks using the kuramoto model,

    F. Sabo and A. Todri-Sanial, “Classonn: Classification with oscillatory neural networks using the kuramoto model,” IEEE, pp. 1–2, 2024

    2024

  18. [25]

    Training energy-based single-layer hopfield and oscillatory networks with un- supervised and supervised algorithms for image classi- fication,

    M. Abernot and A. Todri-Sanial, “Training energy-based single-layer hopfield and oscillatory networks with un- supervised and supervised algorithms for image classi- fication,”Neural Computing and Applications, vol. 35, no. 25, pp. 18 505–18 518, 2023

    2023

  19. [26]

    Deep oscillatory neural network,

    N. R. Rohan, C. Vigneswaran, S. Ghosh, K. Rajendran, A. Gaurav, and V . S. Chakravarthy, “Deep oscillatory neural network,”Scientific Reports, vol. 15, no. 1, p. 40968, 2025. [Online]. Available: https://doi.org/10.1038/s41598-025-24837-4

    2025 doi

  20. [27]

    Artificial kuramoto oscillatory neurons,

    T. Miyato, S. L ¨owe, A. Geiger, and M. Welling, “Artificial kuramoto oscillatory neurons,” 2025. [Online]. Available: https://arxiv.org/abs/2410.13821

    2025 arXiv

  21. [29]

    General oscillator-based Ising-machine models with phase-amplitude dynamics and polynomial interactions,

    L. Sun, M. X. Burns, and M. C. Huang, “General oscillator-based Ising-machine models with phase-amplitude dynamics and polynomial interactions,”Physical Review Applied, vol. 24, no. 4, p. 044035, Oct. 2025. [Online]. Available: https://link.aps.org/doi/10.1103/7s4v-4hs4

    2025 doi

  22. [30]

    Gpu-accelerated simulated oscillator ising/potts machine solving combinatorial optimization problems,

    Y . E. Gonul, C. E. Kayan, I. Mustafazade, N. Kandasamy, and B. Taskin, “Gpu-accelerated simulated oscillator ising/potts machine solving combinatorial optimization problems,” 2025. [Online]. Available: https://arxiv.org/abs/2505.22631

    2025 arXiv

  23. [31]

    Combinatorial optimization using the Lagrange primal-dual dynamics of parametric oscillator networks,

    S. K. Vadlamani, T. P. Xiao, and E. Yablonovitch, “Combinatorial optimization using the Lagrange primal-dual dynamics of parametric oscillator networks,”Physical Review Applied, vol. 21, no. 4, p. 044042, Apr. 2024. [Online]. Available: https://link.aps.org/doi/10.1103/PhysRev...

    2024 doi

  24. [32]

    An fpga-based max-k-cut accelerator exploiting oscillator synchronization model,

    M. K. Bashar, Z. Li, V . Narayanan, and N. Shukla, “An fpga-based max-k-cut accelerator exploiting oscillator synchronization model,” IEEE, pp. 1–8, 2024

    2024

  25. [33]

    A CMOS-compatible oscillation-based VO2 Ising machine solver,

    O. Maher, M. Jim ´enez, C. Delacour, N. Harnack, J. N ´u˜nez, M. J. Avedillo, B. Linares-Barranco, A. Todri-Sanial, G. Indiveri, and S. Karg, “A CMOS-compatible oscillation-based VO2 Ising machine solver,”Nature Communications, vol. 15, no. 1, p. 3334, Apr. 2024. [Online]. Ava...

    2024

  26. [34]

    Simulated bifurcation assisted by thermal fluctuation,

    T. Kanao and H. Goto, “Simulated bifurcation assisted by thermal fluctuation,”Communications Physics, vol. 5, no. 1, p. 153, Jun. 2022. [Online]. Available: https://www.nature.com/articles/s42005-022-00929-9

    2022

  27. [35]

    Com- putational models based on synchronized oscillators for solving combinatorial optimization problems,

    A. Mallick, M. K. Bashar, Z. Lin, and N. Shukla, “Com- putational models based on synchronized oscillators for solving combinatorial optimization problems,”Physical Review Applied, vol. 17, no. 6, p. 064064, 2022

    2022

  28. [36]

    Solving sudoku puzzles by using hopfield neural networks,

    V . Mladenov, P. Karampelas, C. Pavlatos, and E. Zir- intsis, “Solving sudoku puzzles by using hopfield neural networks,”Proc. of ICACM, vol. 11, pp. 174–179, 2011

    2011

  29. [37]

    Kuramoto,Chemical turbulence

    Y . Kuramoto,Chemical turbulence. Springer, 1984

    1984

  30. [38]

    A control theoretic analysis of oscillator ising machines,

    Y . Cheng, M. Khairul Bashar, N. Shukla, and Z. Lin, “A control theoretic analysis of oscillator ising machines,” Chaos: An Interdisciplinary Journal of Nonlinear Sci- ence, vol. 34, no. 7, 2024

    2024

  31. [39]

    py-sudoku,

    J. Sieu, “py-sudoku,” https://github.com/jeffsieu/py- sudoku, 2024

    2024

  32. [40]

    Harnessing dynamics of van der pol oscillators for phase computing with oscillatory neural networks,

    F. Sabo, F. Du, N. Dinc ¸, and A. Todri-Sanial, “Harnessing dynamics of van der pol oscillators for phase computing with oscillatory neural networks,” 2026

    2026

  33. [41]

    An information theoretic study of a duffing oscillator array reservoir computer,

    M. R. E. U. Shougat, X. Li, T. Mollik, and E. Perkins, “An information theoretic study of a duffing oscillator array reservoir computer,”Journal of Computational and Nonlinear Dynamics, vol. 16, no. 8, p. 081004, 2021

    2021

  34. [42]

    Binary classification via spatiotemporal dynamics in reservoir computing rings of fitzhugh–nagumo neurons,

    O. V . Maslennikov, D. S. Shchapin, and V . I. Nekorkin, “Binary classification via spatiotemporal dynamics in reservoir computing rings of fitzhugh–nagumo neurons,” The European Physical Journal Special Topics, vol. 234, no. 15, pp. 4033–4041, 2025

    2025

  35. [43]

    Lxxxviii. on “relaxation-oscillations

    B. van der Pol, “Lxxxviii. on “relaxation-oscillations”,” The London, Edinburgh, and Dublin Philosophical Mag- azine and Journal of Science, vol. 2, no. 11, pp. 978–992, 1926

    1926

  36. [44]

    Impulses and physiological states in theo- retical models of nerve membrane,

    R. FitzHugh, “Impulses and physiological states in theo- retical models of nerve membrane,”Biophysical journal, vol. 1, no. 6, pp. 445–466, 1961

    1961

  37. [45]

    An active pulse transmission line simulating nerve axon,

    J. Nagumo, S. Arimoto, and S. Yoshizawa, “An active pulse transmission line simulating nerve axon,”Proceed- ings of the IRE, vol. 50, no. 10, pp. 2061–2070, 1962

  38. [46]

    Kovacic and M

    I. Kovacic and M. J. Brennan,The Duffing equation: nonlinear oscillators and their behaviour. John Wiley & Sons, 2011

    2011

  39. [47]

    On the stabil- ity and robustness of stuart-landau oscillators,

    E. Panteley, A. Loria, and A. El Ati, “On the stabil- ity and robustness of stuart-landau oscillators,”IFAC- PapersOnLine, vol. 48, no. 11, pp. 645–650, 2015

    2015

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.