Pith. sign in

REVIEW 4 major objections 8 minor 94 references

Ptolemy's Equant Equates to a Universal Dynamical Clock via Machine Learning

T0 review · 4 major / 8 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper claims that for any attractive limit-cycle oscillator there exists a 'dynamical clock' — an equant point and invertible map making the phase a uniformly advancing viewing angle — with closed-form phase dynamics under noise, forcin

desk verdict Genuinely new ML phase-reduction construction spoiled by unsupported off-cycle gradient claims and internal inconsistencies; worth refereeing, not accepting as is. read the letter →

arxiv 2607.15472 v1 pith:BXQXZO7W submitted 2026-07-16 math.DS cs.LGphysics.bio-ph

Jingdong Zhang , Luan Yang , Murilo S. Baptista , Zefeng Zhang , Qunxi Zhu , Wei Lin , Celso Grebogi This is my paper · ORCID
classification math.DScs.LGphysics.bio-ph MSC 34C1537C1037C2737N25
keywords equantdynamicalclockphasereductionlimitcycleresponsecurvearealuniformitycriticaltransitionquorumsensing
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

Oscillations in biology, chemistry, and engineering are routinely described by a 'phase,' but a physically meaningful phase for high-dimensional, irregular limit cycles has remained undefined. This paper claims that every attractive limit-cycle oscillator admits a universal dynamical clock: an equant observation point and an invertible map that make the motion look like uniform rotation on a circle, with the phase equal to the viewing angle from the equant. The clock is built from trajectory data using an autoencoder plus an invertible network, which also yield the phase response curve and, under noise, the curvature correction. If correct, this gives a system-agnostic phase-reduction scheme whose predictions include a superlinear scaling law for E. coli quorum sensing, a route to genetic-circuit modulation, a classical Berry phase, and a geometric early-warning indicator for Hopf bifurcations.

What carries the argument

The equant, defined as the observer point minimizing the normalized areal variance K̂(x) = K(x)/Ā²(x) of swept area over one period (a Kepler-like areal-uniformity criterion). The construction pairs an autoencoder that learns the phase map φ(x) and inverse χ(φ) with dφ/dt = w, and an invertible neural network that extends the circle homeomorphism to a map between the line-of-sight surface and the unit disk, pinning the equant to the disk's center. The phase response curve Z(φ) = ∂φ/∂x|_χ(φ) then carries all perturbation information, yielding closed-form Γ(φ,t) = Z(φ)·P(χ(φ),t); for noise the Hessian Y(φ)=∇²φ enters the Itô drift.

What would settle it

Compare the phase dynamics predicted from the learned φ and ∇²φ against direct simulation of the full stochastic system at increasing noise strengths: if the stationary phase distribution deviates beyond a tolerance while the perturbation stays in the cycle's neighborhood, the unconstrained transversal gradient is the cause. A cheaper deterministic check: on a known model (e.g., FitzHugh–Nagumo), compare the learned Z(φ) against the adjoint-equation PRC computed with the exact Jacobian along the cycle; the paper reports such a comparison in the SM but only for two low-dimensional models, leavi

Watch

Extended reading notes

Core claim

The paper proposes the principle of a universal dynamical clock: for any attractive limit-cycle oscillator, there exists an equant — an observer point selected to minimize the normalized areal variance of the swept area along the cycle — and an invertible map sending lines of sight from the equant to rays from the center of the unit circle, such that the oscillator's phase is the viewing angle from the equant and advances at constant speed w. Once learned from data, the phase map yields the phase response curve Z(φ)=∂φ/∂x, and the perturbed phase dynamics take the closed form dφ/dt=w+Z(φ)·P(χ(φ),t), with a Hessian term entering under noise. The paper demonstrates this clock across ten oscill

Load-bearing premise

The phase map is learned only on the limit cycle itself, and the paper does not constrain how it behaves just off the cycle, yet the reduced dynamics for noise and moderate perturbations rely on those off-cycle gradients being correct.

Editorial extensions

If this is right

  • E. coli quorum-sensing phase-difference dynamics follow a superlinear scaling law in coupling strength Q, answering the 2004 open problem negatively: not Kuramoto-linear.
  • The learned phase response curve lets one predict frequency shifts from static genetic modifications and the phase-lock/phase-slip boundary under periodic forcing, matching electronic-circuit analogues of the Repressilator.
  • The dynamical-clock phase yields a classical-mechanics counterpart of the Berry geometric phase for adiabatic cyclic parameter variation, verified numerically on the FitzHugh–Nagumo system.
  • The optimal equant non-uniformity K̂* decreases toward zero and scales linearly as the FitzHugh–Nagumo parameter approaches a Hopf bifurcation, allowing extrapolated prediction of the critical parameter (predicted ε_c = 1.575 vs true 1.574).
  • Reduced phase dynamics from the clock reproduce synchronization degrees, stationary phase-difference distributions under noise, and phase-lock regions for coupled FitzHugh–Nagumo networks across five real network structures and three coupling types.

Reading between the lines

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

  • If the transversal-extension assumption holds, the same construction should transfer directly to experimental time series from any rhythmic system (neurons, cardiac cells, circadian clocks), since it needs only limit-cycle samples and velocity information.
  • The equant non-uniformity K̂* might serve as a dimension-free geometric order parameter for comparing limit-cycle regularity across systems; one could test whether it correlates with classical measures like Floquet exponents or phase-response amplitude in larger benchmark families.
  • An explicit falsifier of the unregularized method is to compare predicted noise-induced shifts (using learned ∇²φ) against stochastic simulations of the full system for moderate noise; the paper validates on FitzHugh–Nagumo but not on high-dimensional clocks like the 44-dimensional Cdks model.
  • Different equant selection criteria (e.g., minimizing phase-response harmonic content instead of areal variance) could yield alternative 'clocks' with different early-warning and control properties, and the paper's framework could be used to compare them.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 8 minor

Summary. The paper proposes a 'universal dynamical clock' representation for limit-cycle oscillators: an autoencoder learns a phase φ satisfying dφ/dt = w on the limit cycle, and an invertible neural network extends the boundary homeomorphism to a disk, selecting an 'equant' x* by minimizing a normalized areal-variance index K̂. The framework then derives reduced phase dynamics under stochastic, periodic, and network perturbations, and applies it to E. coli quorum-sensing scaling, synthetic genetic-circuit modulation, a classical Berry-phase analogue, and early-warning prediction of a Hopf transition. Validation is reported on ten oscillatory systems, with comparisons to adjoint-based PRCs and extensive supplementary ablations.

Significance. If the claims held, this would be a useful operational phase-reduction toolkit: it uses only limit-cycle trajectories, produces PRCs by autodiff, supplies a geometric observable K̂, and is demonstrated on systems up to 44 dimensions. The supplementary material is detailed and the FitzHugh–Nagumo validations of geometric phase and synchronization are encouraging. However, the stochastic reduction, the E. coli scaling conclusion, and the early-warning claim are not yet supported. The existence of a constant-speed phase coordinate is classical; the novelty lies in the data-driven equant construction and the proposed loss functions. The stress-test concern about transverse-gradient identifiability is real and is the main obstacle to the stochastic phase-reduction claims.

major comments (4)
  1. [§II.2, Eq. (5); §III.1, Eqs. (10)–(14); SM S7; Discussion] The phase map is trained only with L1 and L2 on limit-cycle samples. L1 constrains ∂φ/∂x · F = w, i.e. the tangential derivative of φ on C; it does not constrain the components of ∇φ transverse to F, nor ∇²φ. The PRC Z(φ) = ∇φ|_C is therefore not identified by the data; the learned network has an arbitrary transverse extension. Section III.1's stochastic reduction explicitly uses Y(φ) = ∇²φ|_C in the noise-induced drift and Fokker–Planck equation. The assertion in Sec. II.2 that the learned PRC is 'sufficient' is not proved, and SM S7 introduces adjoint regularization only as an extension, while the Discussion admits that a phase–amplitude extension 'would require off-cycle data or additional regularisation.' Consequently, the E. coli extrinsic-noise results and the stationary phase-difference predictions in Sec. IV.5 rest on an uncontrolled extension.
  2. [§III.1, Eq. (10); SM Eq. (S1.28)] The displayed Fokker–Planck equation has the diffusion term ½Σᵢ∂²φ[(Z·Gᵢ)p]. For an Itô SDE with dφ = (…)dt + Σᵢ Γᵢ(φ)dWᵢ, the correct term is ½Σᵢ∂²φ[Γᵢ(φ)²p]. This is not a notational slip: the stationary distribution and the noise-induced drift calculation in Sec. III.1 depend on the diffusion coefficient. As written, Eq. (10) is mathematically incorrect and must be corrected or the notation must be defined so that the displayed expression is a typographical placeholder.
  3. [§IV.1, Fig. 2J, Eq. (B11); Abstract] The abstract states that E. coli collective oscillations obey a 'super-linear scaling law,' while Sec. IV.1 says NS(Q) 'approximates the linear function y=x at the endpoints of [0,1] but remains below it within the interval, exhibiting second-order scaling.' On [0,1], remaining below y=x is sublinear, not superlinear. The claimed resolution of the 2004 Strogatz problem is therefore internally inconsistent: either the scaling is sublinear (a negative answer to Kuramoto-like linear scaling) or superlinear (a stronger deviation), but not both. The paper should report the fitted scaling and remove the contradiction.
  4. [§IV.4.2, Fig. 5D–E; Table S1; Appendix D] The early-warning claim relies on K̂* decreasing monotonically and approximately linearly as ε approaches the Hopf point. Table S1 shows non-monotonicity (e.g., ε=1.53 gives 1.1765×10⁻⁴, ε=1.54 gives 1.2706×10⁻⁴) and the fitted interval [1.40,1.55] is not fully reported (no 1.55 row). The predicted ε_c = 1.575 is the zero of a linear fit over a selected interval, not a genuine forecast. In addition, Appendix D proves K̂*>0 only for ε=0.05 (εb=0.01), not for the transition regime ε≈1.4–1.57 used in Fig. 5, so the statement that the proof covers 'the considered parameter regime' is inaccurate.
minor comments (8)
  1. [Throughout] Section numbering is inconsistent: the Introduction refers to 'Sec. 1.4' and 'Section 1.4' for the perturbation analysis, but the actual sections are numbered II, III, and IV. Please harmonize the cross-references.
  2. [Abstract; §IV] The abstract lists 'four findings,' but Section IV presents five applications (E. coli, genetic circuit, Berry phase, early warning, and synchronization mechanisms). Please align the counts.
  3. [§IV.2; Appendix C] The genetic-circuit results are obtained from Tina-TI electronic circuit simulations, not from wet-lab experiments. The abstract's phrasing 'experimental data' could be misread; clarify that the data are from circuit simulations.
  4. [§II.2, Eq. (5)] The loss L1 uses a learnable w while minimizing |∂φ/∂x·F − w|²; without a nontriviality constraint, w=0 and constant φ are trivial solutions. The auxiliary L_ptp is introduced later and only for some systems. Consider fixing w by the Fourier frequency or adding the 2π-periodicity constraint to the main objective.
  5. [Definition II.1] The minimization in Eq. (2) is over all x∈R^n, which is unbounded; the existence discussion mentions a bounded admissible domain only in passing. The admissible domain and conditions for existence/uniqueness should be stated explicitly.
  6. [§II.3, L5, and Definition II.2] The line-of-sight loss L5 assumes that the limit cycle is star-shaped with respect to the learned equant, so that the surface M is a well-defined disk. For non-star-shaped limit cycles, M may self-intersect and Definition II.2 may be unsatisfiable. This geometric assumption is not discussed.
  7. [Fig. 5E; Table S1] Provide the fitted linear equation, R², residuals, and the full data in [1.40,1.55] so that the extrapolated ε_c = 1.575 can be reproduced.
  8. [SM S6] The comparison with adjoint PRCs is shown for a single system and without error bars over random seeds. Report variability across seeds and across systems before claiming comparable performance.

Circularity Check

3 steps flagged · score 6.0 of 10

The central 'universal dynamical clock' equivalence is largely definitional: the phase coordinate is assumed/trained via the L1 loss, the equant is defined as the minimizer of the very index later used as evidence, and the validation metrics largely restate the training losses.

  1. self definitional [Section II.1 (Problem setup) and Section II.2, Eq. (5)]
    "We assume there exists a periodic scalar function ϕ(x) on C with period 2π such that its time derivative is a constant w, i.e., dϕ(x)/dt = w ... Such a phase function establishes a homeomorphism between the limit cycle and the unit circle ... L1(θϕ,w) = 1/K Σ_i |∂ϕθϕ/∂x · F(x_i) − w|²₂."

    The main claim—that an arbitrary attractive limit cycle can be equated with a two-dimensional uniform circular motion—is not derived from the dynamics; it is assumed at the outset as the existence of a phase function with constant time derivative, and the autoencoder is trained with L1 to realize precisely this condition. Every periodic orbit admits such a time-parametrized phase (φ=2πt/T), so the 'universal clock' equivalence is a reparameterization guaranteed by construction rather than an empirical discovery. The equant and viewing-angle interpretation are then fitted to match the same phase, making the central equivalence definitional.

  2. self definitional [Definition II.1 and Section IV.4 (Early-warning / K̂*)]
    "Definition II.1 — We define the equant ... x∗ ∈ arg min x K̂(x), K̂(x)=K(x)/Ā²(x) ... We define the optimal equant non-uniformity as K̂∗ = min x K̂(x)."

    The equant is defined as the minimizer of the normalized areal variance K̂, and the paper then presents K̂* as a 'geometric early-warning signal' whose decay near the Hopf bifurcation is used to predict the critical parameter. But K̂* is by definition the value of the paper's own minimization objective, and the existence of a minimizer is already asserted from continuity and boundedness immediately after the definition. The 'prediction' ε_c = 1.575 is obtained by fitting a line to the learned K̂* values and extrapolating to zero; this is an empirical curve fit to the model's own objective, not an independent first-principles prediction. The signal is thus the optimized loss renamed as a physical indicator.

1 more flagged steps
  1. fitted input called prediction [Section IV.4.1 (Performance metrics, 'Universality') and Eq. (5)]
    "Universality. To ensure the existence of the universal dynamical clock, ... we assess the universality of the dynamical clock of each system as, e_i = |w − ∂ϕθϕ/∂x_i · F(x_i)|, i=1,...,K."

    The 'universality' metric used to validate the framework is literally the L1 term of the phase-learning loss in Eq. (5), evaluated on the same limit-cycle training samples. A small e_i is the training objective, not an independent test. Similarly, the 'Uniformity' and 'Interpretability' metrics in the same section check exactly the line-to-line mapping and angle conditions that loss L5 and Definition II.2 were trained/constructed to enforce. The performance section therefore largely reports residuals of the fitted losses, so it cannot independently confirm the existence or universality of the dynamical clock beyond showing that the neural networks minimized their own training objectives.

full rationale

The paper contains substantial independent engineering—neural phase learning, equant optimization, comparisons with adjoint PRCs in the SM, an electronic Repressilator circuit, and detailed dynamical validations—so this is not a case where the entire derivation is vacuous. However, the central theoretical claim that arbitrary limit cycles are 'equated to a universal dynamical clock' is close to definitional: any attractive limit cycle admits a phase coordinate with constant angular speed, and the paper simply trains a network to find such a coordinate (L1) and then defines/optimizes an equant so that the viewing angle coincides with that coordinate. The equant and its non-uniformity K̂* are defined as the solution and minimized value of the paper's own areal-variance objective, and the early-warning 'prediction' is a linear extrapolation of that fitted quantity. The validation metrics for universality, uniformity, and interpretability largely reproduce the training losses. The stochastic and Hessian-related phase reduction (Sec. III.1) is a genuine correctness concern—limit-cycle-only training does not determine the transverse gradient/Hessian of the phase map—but that is a gap in support rather than a circular reduction, and the paper's own Discussion acknowledges that off-cycle data or additional regularization would be needed. Self-citations are present but not load-bearing for the central derivation. Overall, the central equivalence reduces by construction to the paper's own definitions and training objectives, warranting a partial-circularity score of 6 rather than a full 8 or 10, because the ML implementations and several empirical applications retain independent content.

Assumptions & free parameters 6 free parameters · 8 assumptions · 2 invented entities

The central contribution pulls two mathematical structures from the prior literature: phase reduction (isochrons, phase response curves) and universal approximation by neural networks. The genuinely new items are the K̂-based equant selection and the two-stage invertible map, but both are fitted quantities: the equant is the minimizer of a learned loss, the E. coli scaling is read off separately fitted per-Q PRCs, and the early-warning critical point is the zero of a fitted line. The listed free parameters and ad hoc regularizations are load-bearing for those claims.

free parameters (6)
  • Natural frequency w = per system, learned via L1 loss (matched to 2πf in validation)
    Learned as a trainable parameter in the autoencoder loss L1; central to the phase map and the uniform-clock claim.
  • Equant x* = g^{-1}(0) = per system, e.g. FHN ε=1.41: (-0.3450, 0.5554)
    Selected by minimizing normalized areal variance K̂ through the invertible NN; regularized with L2 to avoid a claimed trivial minimizer at infinity; uniqueness is not established.
  • Affine preprocessing (A,b) = per system, up to 44-dimensional (e.g. Cdks b vector length 44)
    Chosen by hand for numerical conditioning; authors argue it is an invertible coordinate change, but it changes the admissible equant search domain and can interact with the L2 regularization.
  • Regularization weights λ1..λ5, L_ptp, L_norm = e.g. L_norm ×1e-4 or 1e-5 applied to some systems only
    Hand-tuned; auxiliary regularizations are applied only for some systems to avoid degenerate solutions, altering the loss landscape and hence the learned equant.
  • Per-density phase functions for E. coli = a set of autoencoders trained on reorganized F_Q for a grid of Q∈[0,1]
    The E. coli scaling law NS(Q) is computed from separately fitted phase-response functions Γ_Q; it is read off per-density fits rather than derived analytically.
  • Early-warning linear fit = line over ε∈[1.40,1.55]; extrapolated ε_c=1.575
    The predicted critical parameter is the zero of a best-fit line through learned K̂* values, not a parameter-free first-principles prediction.
assumptions (8)
  • domain assumption The system has a stable attractive limit cycle and a smooth phase function φ with dφ/dt = w.
    Sec. II.1, Eqs. (1)-(2); standard phase-reduction setting, excludes non-periodic or non-attracting dynamics.
  • domain assumption Perturbations are weak enough that trajectories remain in a neighborhood of the limit cycle (or in the basin of isochrons).
    Sec. II.1 states this requirement; used in all three perturbed-phase derivations and in the E. coli and circuit applications.
  • ad hoc to paper The learned phase map's gradient and Hessian on/off the limit cycle are sufficiently accurate even though transversal gradients are not constrained during training.
    Sec. II.2 asserts sufficiency explicitly; Sec. III.1 uses ∇²φ=Y(φ) for the noise-induced drift, whose accuracy is not validated on high-dimensional systems.
  • ad hoc to paper An invertible extension g maps the line-of-sight surface M to the disk D, sends the equant to the center, and maps line segments to line segments; the trained RealNVP approximates it.
    Sec. II.3; this is the object whose existence is asserted and trained via L3, L4, L5; no theorem guarantees the loss minimum yields the exact geometric map.
  • ad hoc to paper For E. coli, the mean-field coupling can be reorganized into Q-dependent self-dynamics plus a coupling that vanishes at synchrony, and phase reduction can be applied with per-Q phase maps.
    Appendix B, Eq. (B2); this reorganization changes the self-vector field for each Q, so the resulting scaling law is conditional on this modeling choice.
  • domain assumption Time-averaging/slow-phase approximation is valid for periodic forcing and network coupling (ψ is slow over one oscillation period).
    Secs. III.2-III.3 and Appendix A; this is the standard Kuramoto averaging approximation.
  • domain assumption The equant minimization is restricted to a bounded admissible domain; without this and the L2 regularization, existence of the claimed minimizer is not guaranteed.
    Definition II.1 and Supplementary Sec. S2; the L2 regularizer is introduced specifically to avoid a 'trivial minimizer at infinity'.
  • ad hoc to paper Appendix D's proof that K̂*>0 for FitzHugh–Nagumo is assumed to cover the early-warning parameter regime.
    Appendix D fixes ε=0.05, but the early-warning experiment and Table S1 use ε∈[1.40,1.55]; the proof's scope therefore does not match the claimed application.
invented entities (2)
  • Equant x* (optimal observer point)
    purpose: Observation point from which a nonlinear limit cycle is viewed as uniform rotation; selected by minimizing normalized areal variance K̂.
    Defined and computed via ML; no independent measurement or falsifiable handle outside the construction; uniqueness is not established.
  • Universal dynamical clock
    purpose: Canonical 2D uniform rotation on the unit circle that any limit-cycle oscillation is mapped to; provides the 'interpretable phase'.
    Equivalent to the standard phase coordinate; the name is new but the object is a latent circle with constant angular speed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Ptolemy's Equant Equates to a Universal Dynamical Clock via Machine Learning." pith.science (2026). https://pith.science/paper/BXQXZO7W

@misc{pith2026260715472,
  author       = {Pith},
  title        = {Pith review of: Ptolemy's Equant Equates to a Universal Dynamical Clock via Machine Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BXQXZO7W}},
  note         = {Machine review of arXiv:2607.15472}
}
read the original abstract

Oscillatory dynamics arise ubiquitously in nonlinear systems, yet identifying a physically interpretable phase and phase dynamics in nonlinear, high-dimensional oscillations remains a central unresolved problem. Here we establish the principle of a universal dynamical clock, a physical perspective in which oscillations of arbitrary dimensionality and geometry are equivalently represented as uniform rotation through an equant-induced nonlinear viewing coordinate, inspired by Ptolemy's equant and formalised through an areal-uniformity principle reminiscent of Kepler's second law. Using a machine-learning framework, we demonstrate the existence of such an equant for a broad class of oscillatory dynamics and construct the associated dynamical clock and phase dynamics under additive forces, including noise, periodic perturbations, and coupling. Its value in uncovering new physical rules and phenomena is demonstrated by four findings: (i) collective oscillations in Escherichia coli populations obey a previously unexplained superlinear scaling law, resolving a long-standing open problem posed in 2004; (ii) the response mechanisms of engineered genetic circuits to changes in gene expression and environmental conditions; (iii) a classical-mechanics counterpart of the Berry geometric phase emerges naturally from the phase of the dynamical clock; and (iv) optimal equant non-uniformity provides a geometric early-warning signal for critical transitions and enables prediction of critical parameters. By providing operational and system-agnostic phase dynamics that can be constructed directly from data, the dynamical clock enables principled classification, comparison, and control of oscillatory systems, and offers a new route to understanding how specific dynamical regimes support distinct functional behaviours in networked systems.

Figures

Figures reproduced from arXiv: 2607.15472 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

94 extracted references

  1. [1]

    A. L. Hodgkin and A. F. Huxley, A quantitative description of membrane current and its application to conduction and excitation in nerve, The Journal of physiology117, 500 (1952)

    1952

  2. [2]

    A. L. Hodgkin and A. F. Huxley, Propagation of electrical signals along giant nerve fibres, Proceedings of the Royal Society of London. Series B-Biological Sciences140, 177 (1952)

    1952

  3. [3]

    C. Chen, S. Liu, X.-q. Shi, H. Chat´ e, and Y. Wu, Weak synchronization and large-scale collective oscillation in dense bacterial suspensions, Nature542, 210 (2017)

    2017

  4. [4]

    J. A. Brophy and C. A. Voigt, Principles of genetic circuit design, Nat. Meth.11, 508 (2014)

    2014

  5. [5]

    Kuramoto, Chemical turbulence, inChemical oscillations, waves, and turbulence(Springer, Heidelberg, 1984) pp

    Y. Kuramoto, Chemical turbulence, inChemical oscillations, waves, and turbulence(Springer, Heidelberg, 1984) pp. 111– 140

    1984

  6. [6]

    M. S. Yeung and S. H. Strogatz, Time delay in the kuramoto model of coupled oscillators, Physical review letters82, 648 (1999)

    1999

  7. [7]

    Y. B. Kiyohara, M. Katayama, and T. Kondo, A novel mutation in kaic affects resetting of the cyanobacterial circadian clock, Journal of bacteriology187, 2559 (2005)

    2005

  8. [8]

    Ermentrout and A

    B. Ermentrout and A. Mahajan, Simulating, analyzing, and animating dynamical systems: a guide to xppaut for researchers and students, Appl. Mech. Rev.56, B53 (2003)

    2003

Show all 94 references
  1. [9]

    Glass and M

    L. Glass and M. C. Mackey,From clocks to chaos: The rhythms of life(Princeton University Press, Princeton, 1988)

    1988

  2. [10]

    Pietras and A

    B. Pietras and A. Daffertshofer, Network dynamics of coupled oscillators and phase reduction techniques, Phys, Rep.819, 1 (2019)

    2019

  3. [11]

    Evans and C

    J. Evans and C. C. Cannan, Mechanical astronomy: A route to the ancient discovery of epicycles and eccentrics, inFrom Alexandria, through Baghdad: Surveys and studies in the ancient Greek and medieval Islamic mathematical sciences in honor of JL Berggren(Springer, Heidelberg, ...

    2013

  4. [12]

    G. J. Toomer,Ptolemy’s Almagest(Princeton University Press, Princeton, 1998)

    1998

  5. [13]

    R. Iten, T. Metger, H. Wilming, L. Del Rio, and R. Renner, Discovering physical concepts with neural networks, Phys. Rev. Lett.124, 010508 (2020). 54

    2020

  6. [14]

    H. Wang, T. Fu, Y. Du, W. Gao, K. Huang, Z. Liu, P. Chandak, S. Liu, P. Van Katwyk, A. Deac,et al., Scientific discovery in the age of artificial intelligence, Nature620, 47 (2023)

    2023

  7. [15]

    Davies, P

    A. Davies, P. Veliˇ ckovi´ c, L. Buesing, S. Blackwell, D. Zheng, N. Tomaˇ sev, R. Tanburn, P. Battaglia, C. Blundell, A. Juh´ asz, et al., Advancing mathematics by guiding human intuition with ai, Nature600, 70 (2021)

    2021

  8. [16]

    Zhang, Q

    J. Zhang, Q. Zhu, and W. Lin, Learning hamiltonian neural koopman operator and simultaneously sustaining and discov- ering conservation laws, Phys. Rev. Res.6, L012031 (2024)

    2024

  9. [17]

    Zhang, L

    J. Zhang, L. Yang, Q. Zhu, C. Grebogi, and W. Lin, Machine-learning-coined noise induces energy-saving synchrony, Phys. Rev. E110, L012203 (2024)

    2024

  10. [18]

    Monga, D

    B. Monga, D. Wilson, T. Matchen, and J. Moehlis, Phase reduction and phase-based optimal control for biological systems: a tutorial, Biol. Cybern.113, 11 (2019)

    2019

  11. [19]

    A. T. Winfree,The Geometry of Biological Time, Vol. 2 (Springer, 1980)

    1980

  12. [20]

    A. T. Winfree, Patterns of phase compromise in biological cycles, J. Math. Biol.1, 73 (1974)

    1974

  13. [21]

    Guckenheimer, Isochrons and phaseless sets, J

    J. Guckenheimer, Isochrons and phaseless sets, J. Math. Biol.1, 259 (1975)

    1975

  14. [22]

    Brown, J

    E. Brown, J. Moehlis, and P. Holmes, On the phase reduction and response dynamics of neural oscillator populations, Neur. Comput.16, 673 (2004)

    2004

  15. [23]

    D. S. Goldobin, J. Teramae, H. Nakao, and G. B. Ermentrout, Dynamics of limit-cycle oscillators subject to general noise, Phys. Rev. Lett.105, 154101 (2010)

    2010

  16. [24]

    Noviˇ cenko and K

    V. Noviˇ cenko and K. Pyragas, Phase reduction of weakly perturbed limit cycle oscillations in time-delay systems, Phys. D 241, 1090 (2012)

    2012

  17. [25]

    Yawata, K

    K. Yawata, K. Fukami, K. Taira, and H. Nakao, Phase autoencoder for limit-cycle oscillators, Chaos: An Interdisciplinary Journal of Nonlinear Science34(2024)

    2024

  18. [26]

    Hiruta and K

    Y. Hiruta and K. Ishimoto, Autoencoder for limit cycle in kolmogorov flow, Journal of the Physical Society of Japan94, 064401 (2025)

    2025

  19. [27]

    Mauroy and I

    A. Mauroy and I. Mezi´ c, On the use of fourier averages to compute the global isochrons of (quasi) periodic dynamics, Chaos: An Interdisciplinary Journal of Nonlinear Science22(2012)

    2012

  20. [28]

    Mauroy, B

    A. Mauroy, B. Rhoads, J. Moehlis, and I. Mezic, Global isochrons and phase sensitivity of bursting neurons, SIAM Journal on Applied Dynamical Systems13, 306 (2014)

    2014

  21. [29]

    Ermentrout, Type i membranes, phase resetting curves, and synchrony, Neur

    B. Ermentrout, Type i membranes, phase resetting curves, and synchrony, Neur. Comput.8, 979 (1996)

    1996

  22. [30]

    L. Dinh, J. Sohl-Dickstein, and S. Bengio, Density estimation using real nvp, inInternational Conference on Learning Representations(OpenReview, San Juan, Puerto Rico, 2016)

    2016

  23. [31]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga,et al., Pytorch: An imperative style, high-performance deep learning library, inAdvances in Neural Information Processing Systems, Vol. 32 (Curran Associates, Inc., 2019)

    2019

  24. [32]

    Paszke, S

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer, Automatic differentiation in pytorch, NeurIPS 2017 Workshop Autodiff Decision Program Chairs (2017)

    2017

  25. [33]

    Gonze, J

    D. Gonze, J. Halloy, and A. Goldbeter, Robustness of circadian rhythms with respect to molecular noise, Proc. Natl. Acad. Sci.99, 673 (2002)

    2002

  26. [34]

    Yan and A

    J. Yan and A. Goldbeter, Robust synchronization of the cell cycle and the circadian clock through bidirectional coupling, J. Roy. Soc. Interface16, 20190376 (2019)

    2019

  27. [35]

    L. M. Pecora and T. L. Carroll, Master stability functions for synchronized coupled systems, Phys. Rev. Lett.80, 2109 (1998)

    1998

  28. [36]

    Oksendal,Stochastic Differential Equations: An Introduction with Applications(Springer Science & Business Media, 2013)

    B. Oksendal,Stochastic Differential Equations: An Introduction with Applications(Springer Science & Business Media, 2013)

    2013

  29. [37]

    Nakao, K

    H. Nakao, K. Arai, and Y. Kawamura, Noise-induced synchronization and clustering in ensembles of uncoupled limit-cycle oscillators, Phys. Rev. Lett.98, 184101 (2007)

    2007

  30. [38]

    Nakao, Phase reduction approach to synchronisation of nonlinear oscillators, Contemp

    H. Nakao, Phase reduction approach to synchronisation of nonlinear oscillators, Contemp. Phys.57, 188 (2016)

    2016

  31. [39]

    Kuramoto and H

    Y. Kuramoto and H. Nakao, On the concept of dynamical reduction: the case of coupled oscillators, Philos. Trans. Royal Soc. A377, 20190041 (2019)

    2019

  32. [40]

    Garcia-Ojalvo, M

    J. Garcia-Ojalvo, M. B. Elowitz, and S. H. Strogatz, Modeling a synthetic multicellular clock: repressilators coupled by quorum sensing, Proceedings of the National Academy of Sciences101, 10955 (2004)

    2004

  33. [41]

    McMillen, N

    D. McMillen, N. Kopell, J. Hasty, and J. Collins, Synchronizing genetic relaxation oscillators by intercell signaling, Proc. Natl. Acad. Sci.99, 679 (2002)

    2002

  34. [42]

    A. F. Taylor, M. R. Tinsley, F. Wang, Z. Huang, and K. Showalter, Dynamical quorum sensing and synchronization in large populations of chemical oscillators, Science323, 614 (2009)

    2009

  35. [43]

    E. H. Hellen, J. Kurths, and S. K. Dana, Electronic circuit analog of synthetic genetic networks: Revisited, Eur. Phys. J. Spec. Top.226, 1811 (2017)

    2017

  36. [44]

    W. J. Holtz and J. D. Keasling, Engineering static and dynamic control of synthetic pathways, Cell140, 19 (2010)

    2010

  37. [45]

    M. V. Berry, Quantal phase factors accompanying adiabatic changes, Proc. Roy. Soc. A. Math. Phys. Sci.392, 45 (1984)

    1984

  38. [46]

    Simon, Holonomy, the quantum adiabatic theorem, and berry’s phase, Phys

    B. Simon, Holonomy, the quantum adiabatic theorem, and berry’s phase, Phys. Rev. Lett.51, 2167 (1983)

    1983

  39. [47]

    Aharonov and D

    Y. Aharonov and D. Bohm, Significance of electromagnetic potentials in the quantum theory, Phys. Rev.115, 485 (1959)

    1959

  40. [48]

    Pancharatnam, Generalized theory of interference, and its applications: Part i

    S. Pancharatnam, Generalized theory of interference, and its applications: Part i. coherent pencils, inProc. Ind. Acad. Sci. A, Vol. 44 (Springer, 1956) pp. 247–262. 55

    1956

  41. [49]

    J. H. Hannay, Angle variable holonomy in adiabatic excursion of an integrable hamiltonian, J. Phys. A: Math. Gener.18, 221 (1985)

    1985

  42. [50]

    Wilczek and A

    F. Wilczek and A. Shapere,Geometric phases in physics, Vol. 5 (World Scientific, Singapore, 1989)

    1989

  43. [51]

    T. B. Kepler and M. L. Kagan, Geometric phase shifts under adiabatic parameter changes in classical dissipative systems, Phys. Rev. Lett.66, 847 (1991)

    1991

  44. [52]

    S. L. Brunton, J. L. Proctor, and J. N. Kutz, Discovering governing equations from data by sparse identification of nonlinear dynamical systems, Proc. Natl. Acad. Sci.113, 3932 (2016)

    2016

  45. [53]

    Materials and methods are available as supplementary material

  46. [54]

    Rosenblum, A

    M. Rosenblum, A. Pikovsky, J. Kurths, C. Sch¨ afer, and P. A. Tass, Phase synchronization: from theory to data analysis, inHandbook of biological physics, Vol. 4 (Elsevier, Amsterdam, 2001) pp. 279–321

    2001

  47. [55]

    S. H. Strogatz,Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and engineering(CRC press, Florida, 2018)

    2018

  48. [56]

    Varela, J

    F. Varela, J. Lachaux, E. Rodriguez, and J. Martinerie, The brainweb: phase synchronization and large-scale integration, Nat. Rev. Neurosci.2, 229 (2001)

    2001

  49. [57]

    Sauseng and W

    P. Sauseng and W. Klimesch, What does phase information of oscillatory brain activity tell us about cognitive processes?, Neurosci. Biobehav. Rev.32, 1001 (2008)

    2008

  50. [58]

    J. M. Palva, S. Palva, and K. Kaila, Phase synchrony among neuronal oscillations in the human cortex, J. Neurosci.25, 3962 (2005)

    2005

  51. [59]

    X. Li, J. Wang, and W. Hu, Effects of chemical synapses on the enhancement of signal propagation in coupled neurons near the canard regime, Phys. Rev. E76, 041902 (2007)

    2007

  52. [60]

    C. K. Volos, I. M. Kyprianidis, I. N. Stouboulos, E. Tlelo-Cuautle, and S. Vaidyanathan, Memristor: A new concept in synchronization of coupled neuromorphic circuits., J. Eng. Sci. Technol. Rev.8(2015)

    2015

  53. [61]

    A. G. Korotkov, A. O. Kazakov, T. A. Levanova, and G. V. Osipov, The dynamics of ensemble of neuron-like elements with excitatory couplings, Commun. Nonlinear Sci. Numer. Simul.71, 38 (2019)

    2019

  54. [62]

    Hsieh and J

    G.-C. Hsieh and J. C. Hung, Phase-locked loop techniques. a survey, IEEE Trans. Ind. Electron.43, 609 (1996)

    1996

  55. [63]

    Monga and J

    B. Monga and J. Moehlis, Optimal phase control of biological oscillators using augmented phase reduction, Biol. Cybern. 113, 161 (2019)

    2019

  56. [64]

    Zhong, W

    Z. Zhong, W. Lin, and B. Qin, Modulating biological rhythms: A noncomputational strategy harnessing nonlinearity and decoupling frequency and amplitude, Phys. Rev. Lett.131, 138401 (2023)

    2023

  57. [65]

    K. Wang, L. Yang, S. Zhou, and W. Lin, Desynchronizing oscillators coupled in multi-cluster networks through adaptively controlling partial networks, Chaos33(2023)

    2023

  58. [66]

    Kim and D

    J. Kim and D. B. Forger, A mechanism for robust circadian timekeeping via stoichiometric balance, Mol. Syst. Biol.8, 630 (2012)

    2012

  59. [67]

    Shirasaka, W

    S. Shirasaka, W. Kurebayashi, and H. Nakao, Phase reduction theory for hybrid nonlinear oscillators, Physical Review E 95, 012212 (2017)

    2017

  60. [68]

    Wilson and B

    D. Wilson and B. Ermentrout, Greater accuracy and broadened applicability of phase reduction using isostable coordinates, Journal of mathematical biology76, 37 (2018)

    2018

  61. [69]

    Wilson and B

    D. Wilson and B. Ermentrout, Augmented phase reduction of (not so) weakly perturbed coupled oscillators, SIAM Review 61, 277 (2019)

    2019

  62. [70]

    Botvinick-Greenhouse, R

    J. Botvinick-Greenhouse, R. Martin, and Y. Yang, Invariant measures in time-delay coordinates for unique dynamical system identification, Physical Review Letters135, 167202 (2025)

    2025

  63. [71]

    J. A. Acebr´ on, L. L. Bonilla, C. J. P. Vicente, F. Ritort, and R. Spigler, The kuramoto model: A simple paradigm for synchronization phenomena, Rev. Mod. Phys.77, 137 (2005)

    2005

  64. [72]

    J. J. Teo and R. Sarpeshkar, The merging of biological and electronic circuits, Iscience23(2020)

    2020

  65. [73]

    E. H. Hellen, E. Volkov, J. Kurths, and S. K. Dana, An electronic analog of synthetic genetic networks, PLoS One6, e23286 (2011)

    2011

  66. [74]

    M. G. Rosenblum, A. S. Pikovsky, and J. Kurths, Phase synchronization of chaotic oscillators, Phys. Rev. Lett.76, 1804 (1996)

    1996

  67. [75]

    C. M. Gray, Synchronous oscillations in neuronal systems: mechanisms and functions, J. Comput. Neurosci.1, 11 (1994)

    1994

  68. [76]

    Zhang, Q

    J. Zhang, Q. Zhu, and W. Lin, Neural stochastic control, Advances in Neural Information Processing Systems35, 9098 (2022)

    2022

  69. [77]

    FitzHugh, Mathematical models of excitation and propagation in nerve, Biol

    R. FitzHugh, Mathematical models of excitation and propagation in nerve, Biol. Eng.9, 1 (1969)

    1969

  70. [78]

    P. E. Kloeden, E. Platen, and E. Platen,Stochastic differential equations(Springer, 1992)

    1992

  71. [79]

    J. D. Hamilton,Time series analysis(Princeton university press, Princeton, 2020)

    2020

  72. [80]

    Altinok, F

    A. Altinok, F. L´ evi, and A. Goldbeter, A cell cycle automaton model for probing circadian patterns of anticancer drug delivery, Adv. Drug Deliv. Rev59, 1036 (2007)

    2007

  73. [81]

    L´ evi, A

    F. L´ evi, A. Altinok, J. Clairambault, and A. Goldbeter, Implications of circadian clocks for the rhythmic delivery of cancer therapeutics, Philos. Trans. Roy. Soc. A: Math. Phys. Eng. Sci.366, 3575 (2008)

    2008

  74. [82]

    Goldbeter and J

    A. Goldbeter and J. Yan, Multi-synchronization and other patterns of multi-rhythmicity in oscillatory biological systems, Interface Focus12, 20210089 (2022)

    2022

  75. [83]

    Leloup and A

    J. Leloup and A. Goldbeter, Toward a detailed computational model for the mammalian circadian clock, Proc. Natl. Acad. Sci.100, 7051 (2003). 56

    2003

  76. [84]

    Morris and H

    C. Morris and H. Lecar, Voltage oscillations in the barnacle giant muscle fiber, Biophys. J.35, 193 (1981)

    1981

  77. [85]

    H. R. Wilson and J. D. Cowan, Excitatory and inhibitory interactions in localized populations of model neurons, Biophys. J.12, 1 (1972)

    1972

  78. [86]

    E. E. Selkov, Self-oscillations in glycolysis 1. a simple kinetic model, Eur. J. Biochem.4, 79 (1968)

    1968

  79. [87]

    I. M. Bomze, Lotka-volterra equation and replicator dynamics: a two-dimensional classification, Biol. Cybern.48, 201 (1983)

    1983

  80. [88]

    Dolcemascolo, A

    A. Dolcemascolo, A. Miazek, R. Veltz, F. Marino, and S. Barland, Effective low-dimensional dynamics of a mean-field coupled network of slow-fast spiking lasers, Phys. Rev. E101, 052208 (2020)

    2020

  81. [89]

    Goldbeter, A minimal cascade model for the mitotic oscillator involving cyclin and cdc2 kinase., Proc

    A. Goldbeter, A minimal cascade model for the mitotic oscillator involving cyclin and cdc2 kinase., Proc. Natl. Acad. Sci. 88, 9107 (1991)

    1991

  82. [90]

    J. E. Rubin and D. Terman, High frequency stimulation of the subthalamic nucleus eliminates pathological thalamic rhythmicity in a computational model, J. Comput. Neurosci.16, 211 (2004)

    2004

  83. [91]

    G´ erard and A

    C. G´ erard and A. Goldbeter, A skeleton model for the network of cyclin-dependent kinases driving the mammalian cell cycle, Interface Focus1, 24 (2011)

    2011

  84. [92]

    G´ erard and A

    C. G´ erard and A. Goldbeter, Temporal self-organization of the cyclin/cdk network driving the mammalian cell cycle, Proc. Natl. Acad. Sci.106, 21643 (2009)

    2009

  85. [93]

    R. T. Q. Chen, Y. Rubanova, J. Bettencourt, and D. K. Duvenaud, Neural ordinary differential equations, Advances in neural information processing systems31(2018)

    2018

  86. [94]

    Pathak, B

    J. Pathak, B. Hunt, M. Girvan, Z. Lu, and E. Ott, Model-free prediction of large spatiotemporally chaotic systems from data: A reservoir computing approach, Phys. Rev. Lett.120, 024102 (2018)

    2018

Pith tools

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