Pith. sign in

REVIEW 2 major objections 6 minor 63 references

Synchronization, Kinematic Waves and Spike-Phase-Separation in Feedback Ising Neural Networks on Heterogeneous Graphs

T0 review · 2 major / 6 minor · reviewed 2026-07-31 · grok-4.5

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

Pith's one-line read Degree heterogeneity sets when feedback Ising networks start oscillating, and extreme heterogeneity can kill the synchronized cycle by locking the system into a phase-separated state.

desk verdict Clean closed-form Hopf line on heterogeneous graphs plus two real new phenomena (kinematic waves, m=0/u≠0 phase separation); annealed HMF is the limit, not a break. read the letter →

arxiv 2607.28275 v1 pith:ON5BIHUB submitted 2026-07-30 cond-mat.stat-mech physics.bio-ph

Anna Poggialini , Irem Topal , Fabrizio Lombardi , Daniele De Martino This is my paper · ORCID
classification cond-mat.stat-mechphysics.bio-ph PACS 05.70.Ln64.60.aq87.19.lj05.45.-a
keywords feedbackIsingmodelheterogeneousmean-fieldAndronov-Hopfbifurcationkinematicwavesphaseseparationdegreeheterogeneityneuralsynchronizationrandomgraphs
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks how uneven connectivity changes collective oscillations in a simple neural-network model: binary spins on a random graph whose excitability is homeostatically pushed against the population firing rate. Using a heterogeneous mean-field theory checked against Monte Carlo runs, the authors show that the silent state loses stability through a Hopf bifurcation whose threshold is fixed by the single number α = ⟨k²⟩/⟨k⟩. Because heterogeneity splits the ordinary magnetization m from a synapse-weighted magnetization u, two effects appear that never occur on regular lattices: degree-ordered kinematic waves that sweep from leaves to hubs, and a low-temperature phase-separated fixed point with m = 0 but u ≠ 0. For sufficiently broad degree distributions that fixed point stabilizes and swallows the limit cycle. The result supplies closed-form phase boundaries that let network topology, not just feedback gain, be read as a control knob for macroscopic neural rhythms.

What carries the argument

The Curie–Weiss heterogeneous mean-field reduction that tracks two distinct order parameters—firing rate per neuron m and firing rate per synapse u—together with the closed-form Hopf line β_c(c) = (3α − c − √((c + α)² − 8c))/(2(α² − c(α − 1))) obtained from the Routh–Hurwitz conditions on their joint Jacobian.

What would settle it

On a large quenched scale-free or bimodal graph, measure whether the analytic Hopf boundary still matches the onset of global oscillations and whether the phase-separated fixed point really captures and extinguishes the limit cycle once degree variance exceeds the predicted threshold.

Watch

Extended reading notes

Core claim

On heterogeneous random graphs the paramagnetic fixed point of the feedback Ising model loses stability via a supercritical Andronov–Hopf bifurcation at the explicit threshold β_c(c) determined solely by the moment ratio α = ⟨k²⟩/⟨k⟩ and the feedback strength c; when degree variance is large enough, a pitchfork-born phase-separated state (m = 0, u ≠ 0) later stabilizes and dynamically destroys the synchronized limit cycle.

Load-bearing premise

The annealed mean-field closure that every node of the same degree shares one magnetization and feels only a single average neighbor field, turning the interaction into a rank-one object.

Editorial extensions

If this is right

  • Scale-free networks with 2 < γ ≤ 3 sit permanently in the oscillatory phase at any nonzero temperature in the thermodynamic limit, independent of feedback strength.
  • Kinematic waves—low-degree nodes leading, hubs lagging—appear both on the limit cycle and as measurable peak delays in subcritical cross-correlations ⟨m_k(t+τ)h(t)⟩.
  • For bimodal graphs the phase-separated point stabilizes exactly when an effective inverse temperature ϕ = β sech²(β Δk u_s/2) drops below the paramagnetic Hopf line, giving a sharp topology-dependent death of synchrony.
  • Slow drift of α (for example through gradual loss of hubs) can push a network across the phase boundary without any change in the feedback gain c.

Reading between the lines

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

  • If cortical degree distributions broaden or narrow with aging or disease, the same homeostatic feedback could flip a circuit from rhythmic to phase-separated silence without altering cellular excitability parameters.
  • The rank-one dark modes that sit at eigenvalue −1 under the annealed approximation should split on quenched graphs, potentially producing additional slow kinematic modes visible in degree-resolved spectra.
  • Extending the same m–u decoupling to non-reciprocal or multi-state spins would likely generate coexisting avalanche and oscillatory regimes whose boundaries still collapse onto α.
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

2 major / 6 minor

Summary. The manuscript studies a kinetic Ising model with homeostatic linear feedback from global magnetization to the external field on random graphs with heterogeneous degree distributions. Within a Curie–Weiss heterogeneous mean-field closure, the authors linearize the (m,u,h) dynamics about the paramagnetic fixed point, apply Routh–Hurwitz, and obtain a closed-form Andronov–Hopf threshold β_c(c) controlled by the moment ratio α=⟨k²⟩/⟨k⟩ (Eq. 19), with limits recovering the known ferromagnetic HMF critical point. Degree heterogeneity decouples the per-neuron rate m from the per-synapse rate u, producing (i) kinematic waves of degree-ordered activation, quantified via cross-correlations C_k(τ), and (ii) a low-temperature phase-separated fixed point (m=0, u≠0) born in a supercritical pitchfork at β_c2=⟨k⟩/Var(k). For bimodal graphs they reduce PS stability exactly to a rescaled paramagnetic condition φ=β sech²(·)=β_Hopf(c). Analytical phase boundaries are compared to Monte Carlo on ER, scale-free, log-normal, and bimodal graphs.

Significance. If the results hold, the paper supplies a rare analytically tractable out-of-equilibrium phase diagram for feedback-driven Ising dynamics on heterogeneous networks, with an explicit topology dependence through α and two phenomena (kinematic waves, spike-phase separation) absent on regular graphs. The closed-form Hopf line, the bimodal reduction φ→β_Hopf, and the van Kampen derivation of the HMF equations are concrete technical contributions that can be reused and falsified. Monte Carlo agreement on several degree ensembles strengthens the claim within the annealed scope. The work cleanly extends prior homogeneous feedback-Ising results and connects network heterogeneity to oscillatory onset and limit-cycle destruction in a way that is useful for statistical mechanics of neural and complex systems.

major comments (2)
  1. [§II.C, Abstract] Abstract and §II.C.5–6 claim that when the phase-separated fixed point stabilizes it “dynamically destroys the synchronized limit cycle.” Stabilization of the PS point is controlled analytically (especially the exact bimodal reduction φ(β_stab)=β_Hopf(c)), but the global mechanism that eliminates the LC (e.g., collision, basin capture, or fold of cycles) is not charted. Fig. 4 shows a quench into PS, which is consistent with capture but does not establish how the LC ceases to exist as a stable object. A minimal addition—continuation of the LC in β for a bimodal or high-variance case, or a clear statement that destruction is inferred from MC basins rather than proven—would make the central claim load-bearing and precise.
  2. [§II.A.2, Eqs. (11)–(19)] The Hopf bifurcation is repeatedly called supercritical (Abstract, Introduction, §II.A), yet the first Lyapunov coefficient is not recomputed for the heterogeneous (m,u,h) Jacobian. Supercriticality is inherited from the homogeneous/lattice feedback-Ising literature. Because the phase diagram and the interpretation of a stable macroscopic limit cycle rest on this, either a short Lyapunov-coefficient calculation (or normal-form reduction) for Eqs. (11)–(13), or an explicit caveat that supercriticality is assumed by continuity with the homogeneous case and supported by MC, should be added.
minor comments (6)
  1. [Appendix B] Appendix B phase-detection thresholds (peak position >0.05, p(0)<1) are ad hoc. State that they were fixed from visual inspection and, if possible, show that the extracted β_c(c) is stable under modest threshold changes, or mark the MC phase boundary as semi-quantitative.
  2. [Figure 2] Fig. 2 insets confirm oscillatory vs fixed-point regimes but do not overlay the analytical β_c(c) on a quantitative MC scan (e.g., oscillation amplitude or spectral peak vs β). A single panel with measured onset vs Eq. (19) would strengthen the validation claim.
  3. [§II.A.2–3] Notation: α is introduced as ⟨k²⟩/⟨k⟩ in the text after Eq. (15), while z is used for ⟨k⟩; keep this consistent in the ER specialization (Eq. 22) and in the PS section where z reappears.
  4. [§II.B, §III] In §II.B the kinematic-wave interpretation is clear, but the claim that the structure is “encoded in the Hopf eigenvector through q_u/q_h=β_H/(α+iω_0)” appears only in the Conclusion. Either move a one-line derivation into §II.B or soften the Conclusion wording.
  5. [Throughout / Appendix A–B] Typos/style: “F eedback”, “W eiss”, “RESUL TS” (space in headings); “withe noise” in Appendix A; “as done done” in Appendix B; arXiv stamp date “30 Jul 2026” looks like a placeholder.
  6. [§II, Eq. (4) and Eqs. (9)–(10)] Briefly clarify the feedback update convention Δt=1/N (after each spin flip) versus the continuous ˙h=−c m used in HMF, so that MC and theory timescales are unambiguously matched.

Circularity Check

1 steps flagged · score 1.0 of 10

No significant circularity: Hopf threshold, m–u decoupling, kinematic waves, and PS pitchfork are derived from the HMF Jacobian/Routh–Hurwitz and checked by independent MC; prior self-cites only supply the homogeneous feedback setup.

  1. self citation load bearing [Sec. I / opening of Sec. II (feedback setup; cites [13–15])]
    "A complementary approach, recently introduced in [13–15], equips Ising-type spin models with a linear feedback between the order parameter (magnetization) and the control parameter (external field). On fully connected graphs and regular lattices, this mechanism replaces the equilibrium ferromagnetic transition with an out-of-equilibrium synchronization transition: the paramagnetic fixed point loses stability via a supercritical Andronov–Hopf bifurcation, giving rise to a macroscopic limit cycle [13–15]."

    The out-of-equilibrium feedback rule and the homogeneous Hopf scenario are imported from overlapping-author prior work rather than re-derived from first principles here. This is ordinary model inheritance, not a reduction of the paper’s new claims: the heterogeneous Jacobian, closed-form β_c(c;α), kinematic-wave correlators, and PS pitchfork/stabilization are derived and MC-tested in this manuscript. Flagged only as minor non-load-bearing self-citation.

full rationale

The load-bearing analytical claims are obtained inside this paper from a stated Curie–Weiss closure, not by renaming a fit or equating a claim to its input. Starting from Glauber rates plus the linear feedback ḣ=−c m, the authors write the degree-class ODEs (10), linearize about the paramagnetic point to get the (m,u,h) Jacobian (11)–(13), form the cubic characteristic polynomial (14)–(15), and extract the Hopf boundary by Routh–Hurwitz as the closed form β_c(c) in Eq. (19) depending only on α=⟨k²⟩/⟨k⟩. The phase-separated fixed point is obtained from the steady-state conditions m=0 and the cubic expansion yielding the supercritical pitchfork at β_c2=⟨k⟩/Var(k) (34)–(35); stability follows from the same Jacobian structure with sech²-weighted moments, and for bimodal graphs reduces exactly to φ(β_stab)=β_Hopf(c) with φ=β sech²(β Δk u_s/2). Kinematic waves are read off as degree-ordered peaks of the correlator C_k(τ), not fitted. Phase diagrams are validated against Monte Carlo on ER, scale-free, log-normal, and bimodal graphs—external to the analytic expressions. Self-citations [13–15] introduce the feedback Ising mechanism on homogeneous graphs/lattices; they are not used as uniqueness theorems that force the heterogeneous results, nor is any free parameter fitted to data and then re-presented as a prediction. Residual modeling choices (annealed rank-one HMF, experimental Hilbert thresholds in App. B for numerical phase maps) are assumptions or post-processing, not circular reductions of the central formulas. Score 1 only for ordinary self-citation of the homogeneous baseline, which is not load-bearing for the new α-dependent and PS claims.

Assumptions & free parameters 3 free parameters · 6 assumptions · 2 invented entities

The load-bearing content is a dynamical mean-field theory on annealed heterogeneous graphs plus a linear homeostatic feedback. Almost all structure comes from standard Glauber Ising dynamics, classical HMF moment closure, and the authors’ previously introduced feedback rule; free parameters are model knobs scanned in phase diagrams rather than fits to external neural data. No new physical particles are postulated—‘kinematic waves’ and ‘spike-phase-separation’ name emergent regimes of the same ODEs.

free parameters (3)
  • feedback strength c = order 0.1 in simulations; asymptotic c→0 and c→∞ analyzed
    Single gain in ḣ=−c m; scanned as a control axis in phase diagrams, not fitted to empirical brain data.
  • degree-distribution parameters (z, γ, k_min, σ, bimodal k1/k2) = e.g. ER z=4 or 20; SF γ=2.1, k_min=20; log-normal ⟨k⟩=25, σ=1
    Choose the graph ensemble and thus α and Var(k); set by hand for comparability across ER/SF/log-normal/bimodal examples.
  • phase-detection KDE thresholds (peak>0.05, p(0)<1) = peak threshold=0.05; p(0) threshold=1
    Appendix B states thresholds are ‘experimentally derived from the signals observation’ for automated (c,β) maps—ad hoc classifiers, not physical constants.
assumptions (6)
  • domain assumption Glauber single-spin flip rates with detailed balance w.r.t. Ising Hamiltonian at fixed h
    Eqs. (2)–(3); standard kinetic Ising setup underlying the master equation in Appendix A.
  • domain assumption Linear homeostatic feedback ḣ=−c m (or Δh=−c m Δt after each flip)
    Eq. (4) and (9); imported from the authors’ prior feedback-Ising framework [13–15], not derived here.
  • ad hoc to paper Heterogeneous Curie–Weiss closure: m_k identical within degree class; neighbor field u=∑ k p_k m_k/⟨k⟩; annealed rank-one interactions
    Sec. II.A.1 Eqs. (5)–(10); enables the 3D (m,u,h) reduction and closed-form Hopf line; acknowledged to create artificial dark-mode degeneracy.
  • standard math Van Kampen system-size expansion: deterministic HMF ODEs obtained by dropping 1/√N_k noise terms
    Appendix A; standard passage from master equation to macroscopic rate equations used for bifurcation analysis.
  • standard math Routh–Hurwitz criterion on the cubic characteristic polynomial determines local stability and Hopf onset
    Sec. II.A.2 Eqs. (14)–(19); classical linear stability tool.
  • domain assumption Random-graph degree ensembles (ER, power-law with structural cutoff, log-normal, bimodal) adequately represent relevant heterogeneity
    Sec. II.A.3–4 and simulation figures; standard complex-network modeling choice.
invented entities (2)
  • spike-phase-separated (PS) fixed point with m=0 but u≠0
    purpose: Name and organize the low-T pitchfork state where degree classes split by magnetization sign at a boundary near the median degree.
    Emergent fixed point of the same HMF ODEs (Eqs. 30–35), not an extra field or particle; independent_evidence false beyond this model class.
  • kinematic waves of degree-ordered activation
    purpose: Describe sequential periphery-to-hub response encoded in Hopf eigenvector / C_k(τ) peak delays.
    Interpretive label (Murray-style kinematic wave) for graded susceptibility across degree classes; no new dynamical variable introduced.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Synchronization, Kinematic Waves and Spike-Phase-Separation in Feedback Ising Neural Networks on Heterogeneous Graphs." pith.science (2026). https://pith.science/paper/ON5BIHUB

@misc{pith2026260728275,
  author       = {Pith},
  title        = {Pith review of: Synchronization, Kinematic Waves and Spike-Phase-Separation in Feedback Ising Neural Networks on Heterogeneous Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ON5BIHUB}},
  note         = {Machine review of arXiv:2607.28275}
}
read the original abstract

Structural heterogeneity constrains collective dynamics in complex systems. However, its analytical tractability out of equilibrium remains limited. In this work, we study a class of kinetic Ising neural networks driven out of equilibrium by a homeostatic feedback loop between the neuronal excitability and the population firing rate. Using a Curie-Weiss heterogeneous mean-field approximation validated by Monte Carlo simulations, we provide an analytical characterization of how a macroscopic synchronized limit cycle emerges via an Andronov-Hopf bifurcation on heterogeneous networks. We derive closed-form phase boundaries and show that the onset of oscillations is explicitly controlled by network heterogeneity through the degree moment ratio. Degree heterogeneity decouples the spiking rate per neuron m from the spiking rate per synapse u, generating physical phenomena absent in homogeneous systems. These include (i) kinematic waves of sequential, degree-ordered activations propagating from the network periphery to the hubs, and (ii) a low-temperature phase-separated state emerging via a pitchfork bifurcation. We prove that for highly heterogeneous topologies, this phase-separated fixed point stabilizes and dynamically destroys the synchronized limit cycle. These results provide a mathematical framework for understanding how heterogeneity regulates macroscopic oscillations and out-of-equilibrium transitions in neural networks

Figures

Figures reproduced from arXiv: 2607.28275 by the authors.

Figure 1
Figure 1. Feedback Ising model on a heterogeneous network. A. Microscopic spin configurations (si = +1 in blue, si = −1 in red) at four representative times along the limit cycle. B. Time series of the global magnetization m(t) and the feedback field h(t), showing nonlinear self-sustained oscillations. Dashed lines mark the snapshots of panel A. C. Degree-resolved magnetizations mk(t) for different degree classes, illustratin… view at source ↗
Figure 2
Figure 2. Analytical phase diagrams in the (c, β) plane from the Curie–Weiss linear stability analysis. Left: Erd˝os–R´enyi graph with z = 20. Center: Scale-free network with exponent γ = 2.1 and kmin = 20. Right: Log-normal degree distribution with ⟨k⟩ = 25 and σ = 1 (µ = ln⟨k⟩ − σ 2/2). Insets: Phase-plane trajectories (m, h) from Monte Carlo simulations confirm the analytical predictions. For z = 20 used in our numerical e… view at source ↗
Figure 3
Figure 3. Kinematic wave in the subcritical resonant phase. Microscopic numerical simulations on a Poissonian network with z = 4, c = 0.1, β = 0.24, N = 104 . Top left: Time series of the external field h and degree-resolved magnetizations m1, m3, m5, showing noisy oscillations with degree-dependent amplitude. Top right: Stationary cross-correlation Ck(τ ) = ⟨mk(t + τ ) h(t)⟩ for k = 1, . . . , 5. The peak amplitude grows wit… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Phase separation after a quench. Microscopic numerical simulations on a Bimodal network with k1 = 4, k2 = 12 c = 0.1, N = 104 . Left: Snapshot of the system after the quench. Right: Magnetization after and before the quench. where the effective moments are an ≡ X k pk …
Figure 5
Figure 5. Figure 5: Computational derivation of the phase diagrams in the (c, β) plane for the Erd˝os–R´enyi graph. Left: The purple and the green areas respectively correspond to the (m = 0, h = 0) fixed point and the limit cycle detection according to the pseudo-code above. Center: To p…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 1 canonical work pages

  1. [1]

    The effec- tive field experienced by a degree-knode depends on the neighbor magnetizationu, defined as the mean magne- tization encountered when following a randomly chosen edge

    Heterogeneous mean-field equations In the heterogeneous mean-field (HMF) approxima- tion, all nodes sharing the same degreekare assumed to have the same average magnetizationm k. The effec- tive field experienced by a degree-knode depends on the neighbor magnetizationu, defined as the mean magne- tization encountered when following a randomly chosen edge....

  2. [2]

    Linear stability analysis and bifurcation condition The paramagnetic statem k =u=h= 0 is always a fixed point of Eqs. (10). To determine the onset of os- cillations, we linearize Eqs. (10) around this fixed point. Writingm= P k pk mk for the global magnetization per- turbation, the linearized system in the variables (m, u, h) reads ˙m=−m+β(⟨k⟩u+h),(11) ˙u...

  3. [3]

    Substituting this into Eq

    Application to Erd˝ os–R´ enyi graphs For a Poissonian (Erd˝ os–R´ enyi) graph with mean de- greez, one has⟨k⟩=z,⟨k 2⟩=z(z+ 1) and therefore α=z+ 1. Substituting this into Eq. (19) gives the ex- plicit form βc(c) = 3(z+ 1)−c− p (c+z+ 1) 2 −8c 2 (z+ 1) 2 −c z ,(22) with limits βc(0) = 1 z+ 1 , β c(∞) = 1 z .(23) 5 a) b) c) Figure 2:Analytical phase diagram...

  4. [4]

    The relevant moments are ⟨k⟩=    γ−1 γ−2 kmin, γ >2, ∼logN, γ= 2, (24) ⟨k2⟩=    γ−1 γ−3 k2 min, γ >3, ∼logN, γ= 3, ∼N (3−γ)/2,2< γ <3 (structural cutoff)

    Extension to strongly heterogeneous networks: power law and log-normal) Next, we consider graphs with power-law degree distri- butionsp(k)∝k −γ withk≥k min, where the maximum degree is set by a structural cutoffk max ∼N 1/2 (or, al- ternatively, the natural cutoffk max ∼N 1/(γ−1)). The relevant moments are ⟨k⟩=    γ−1 γ−2 kmin, γ >2, ∼logN, γ= 2, (24) ...

  5. [5]

    phase-separated

    Fixed-point equations In the steady state, ˙h= 0 impliesm= 0. Together with the self-consistency condition onu, the fixed-point equations read X k pk tanh β(k u+h) = 0,(30) u= X k k pk ⟨k⟩ tanh β(k u+h) .(31) The trivial solutionu=h= 0 (the paramagnetic state) always satisfies these equations. We now ask when a nontrivial “phase-separated” solution withu̸...

  6. [6]

    (30) gives at linear orderh=−⟨k⟩u≡ −z u

    Third-order expansion and bifurcation point Expanding tanh(βx) =βx− 1 3 β3x3 +O(x 5) withx k = ku+h, Eq. (30) gives at linear orderh=−⟨k⟩u≡ −z u. Writingh=−z u+δand substituting back, the cubic correction yields δ= β2 µ3 3 u3 +O(u 5),(32) whereµ n =⟨(k−z) n⟩denotes then-th central moment of the degree distribution. Note that for symmetric dis- tributions ...

  7. [7]

    (30) becomes − X k<k∗ pk + X k>k∗ pk = 0,(36) so thatk ∗ is themedianof the degree distribution

    Zero-temperature limit In the limitβ→ ∞, tanh→sgn and Eq. (30) becomes − X k<k∗ pk + X k>k∗ pk = 0,(36) so thatk ∗ is themedianof the degree distribution. The corresponding solution is us = 1 ⟨k⟩ X k>k∗ k pk − X k<k∗ k pk ! , h s =−k ∗ us . (37) In this limit, the nodes orientation splits into two groups, where nodes of degreek > k∗ andk < k∗ take opposit...

  8. [8]

    Stability of the phase-separated fixed point The Jacobian of the full dynamical system (10) evalu- ated at the phase-separated (PS) point (m= 0, u s, hs) is JPS =   −1β a 1 β a0 0−1 + β a2 z β a1 z −c0 0   ,(38) 8 -1 -0.5 0 0.5 1 9000 9500 10000 10500 11000 mk1=2 mk2=12 quench β= 0.2->1 magnetization time -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 9000 950...

Show all 63 references
  1. [9]

    Stability within the third-order expansion Substituting sech2(βxk) = 1−β 2(k−z) 2u2 s +O(u 4 s) and the expression (35) foru 2 s, the effective moments become a0 = 1− 3µ2(βµ2 −z) β µ4 ,(43) a1 =z− 3(µ3 +zµ 2)(βµ2 −z) β µ4 ,(44) a2 =⟨k 2⟩ −3(µ4 + 2zµ3 +z 2µ2)(βµ2 −z) β µ4 .(45)...

  2. [10]

    Since z2 > µ2 and the denominator is positive for all distri- butions we have examined,ϵ stab >0 and the PS point always stabilizes at a temperature below its birth

    Expanding inϵ=β−β c2 we find A(ϵ) = µ2 −z 2 µ2| {z } A0 <0 + 3µ2 2z2 + 6µ2µ3z+ (2µ 2 −z 2)µ4 z µ4 ϵ+O(ϵ 2).(46) SettingA= 0 yields the stabilization threshold ϵstab =β stab −β c2 = (z2 −µ 2)z µ4 µ2 D , (47) whereD ≡3µ 2 2z2 + 6µ 2µ3z+ (2µ 2 −z 2)µ4. Since z2 > µ2 and the denom...

  3. [11]

    dark modes

    Bimodal random graph: exact finite-βsolution For a bimodal graph with degreesk 1 < k2,p 1 =p 2 = 1 2 , andz= (k 1 +k 2)/2, the phase-separated fixed point can be solved in closed form at arbitraryβ. Them= 0 condition (30) requires tanh(β(k 1u+h)) + tanh(β(k 2u+ h)) = 0, which ...

  4. [12]

    Buend ´ ıa, P

    V. Buend ´ ıa, P. Villegas, R. Burioni, and M. A. Mu˜ noz, The broad edge of synchronization: Griffiths ef- fects and collective phenomena in brain networks, Philosophical Transactions of the Royal Society A: Math- ematical, Physical and Engineering Sciences380, 20200424 (2022...

    2022

  5. [13]

    Bullmore and O

    E. Bullmore and O. Sporns, Complex brain networks: graph theoretical analysis of structural and functional systems, Nature Reviews Neuroscience10, 186 (2009)

    2009

  6. [14]

    Sporns,Networks of the Brain(MIT Press, 2011)

    O. Sporns,Networks of the Brain(MIT Press, 2011)

    2011

  7. [15]

    could show noise-induced switching between the limit cycle and the paramagnetic or phase-separated states in the coexistence regime near the Bautin point. Verifying this prediction in simulations would connect the bifur- cation analysis to the statistics of intermittent oscill...

    2024

  8. [16]

    J. J. Hopfield, Neural networks and physical systems with emergent collective computational abilities, Proceedings of the National Academy of Sciences79, 2554 (1982)

    1982

  9. [17]

    D. J. Amit, H. Gutfreund, and H. Sompolinsky, Spin-glass models of neural networks, Physical Review A32, 1007 (1985)

    1985

  10. [18]

    Buzs´ aki,Rhythms of the Brain(Oxford University Press, 2006)

    G. Buzs´ aki,Rhythms of the Brain(Oxford University Press, 2006)

    2006

  11. [19]

    van Vreeswijk and H

    C. van Vreeswijk and H. Sompolinsky, Chaos in neuronal networks with balanced excitatory and inhibitory activity, Science 274, 1724 (1996)

    1996

  12. [20]

    Brunel, Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons, Journal of Computational Neuroscience8, 183 (2000)

    N. Brunel, Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons, Journal of Computational Neuroscience8, 183 (2000)

    2000

  13. [21]

    B¨ orgers and N

    C. B¨ orgers and N. J. Kopell, Effects of noisy drive on rhythms in networks of excitatory and inhibitory neurons, Neural Computation17, 557 (2005)

    2005

  14. [22]

    H. R. Wilson and J. D. Cowan, Excitatory and inhibitory interactions in localized populations of model neurons, Biophysical Journal12, 1 (1972)

    1972

  15. [23]

    Kuramoto,Chemical Oscillations, Waves, and Turbulence(Springer, 1984)

    Y. Kuramoto,Chemical Oscillations, Waves, and Turbulence(Springer, 1984)

    1984

  16. [24]

    J. A. Acebr´ on, L. L. Bonilla, C. J. P´ erez Vicente, F. Ritort, and R. Spigler, The Kuramoto model: A simple paradigm for synchronization phenomena, Reviews of Modern Physics77, 137 (2005)

    2005

  17. [25]

    De Martino, Feedback-induced self-oscillations in large interacting systems subjected to phase transitions, Journal of Physics A: Mathematical and Theoretical52, 045002 (2019)

    D. De Martino, Feedback-induced self-oscillations in large interacting systems subjected to phase transitions, Journal of Physics A: Mathematical and Theoretical52, 045002 (2019)

    2019

  18. [26]

    De Martino and A

    D. De Martino and A. C. Barato, Oscillations in feedback-driven systems: Thermodynamics and noise, Physical Review E100, 062123 (2019)

    2019

  19. [27]

    Sinelshchikov, A

    D. Sinelshchikov, A. Poggialini, M. F. Abbate, and D. De Martino, Emergence of collective self-oscillations in minimal lattice models with feedback, Physical Review E108, 044204 (2023). 14

    2023

  20. [28]

    Lombardi, S

    F. Lombardi, S. Pepi´ c, O. Shriki, G. Tkaˇ cik, and D. De Martino, Statistical modeling of adaptive neural networks explains co-existence of avalanches and oscillations in resting human brain, Nature Computational Science3, 254 (2023)

    2023

  21. [29]

    Dahmen, A

    D. Dahmen, A. Hutt, G. Indiveri, A. Kennedy, J. Lefebvre, L. Mazzucato, A. E. Motter, R. Narayanan, M. Pay- vand, H. Planert, and R. Gast, How heterogeneity shapes dynamics and computation in the brain, Neuron 10.1016/j.neuron.2025.11.023 (2025), online ahead of print

    2025 doi

  22. [30]

    Albert and A.-L

    R. Albert and A.-L. Barab´ asi, Statistical mechanics of complex networks, Reviews of Modern Physics74, 47 (2002)

    2002

  23. [31]

    M. E. J. Newman, The structure and function of complex networks, SIAM Review45, 167 (2003)

    2003

  24. [32]

    Alternative choices of the feedback timescale affect the entropy production rate and depend on the amount of information available to implement the control loop (here assumed to be complete), but lead to qualitatively similar dynamics

  25. [33]

    Leone, A

    M. Leone, A. V´ azquez, A. Vespignani, and R. Zecchina, Ferromagnetic ordering in graphs with arbitrary degree distribution, The European Physical Journal B-Condensed Matter and Complex Systems28, 191 (2002)

    2002

  26. [34]

    J. D. Murray,Mathematical Biology I: An Introduction, 3rd ed. (Springer, 2002)

    2002

  27. [35]

    S. N. Dorogovtsev, A. V. Goltsev, and J. F. F. Mendes, Ising model on networks with an arbitrary distribution of connec- tions, Physical Review E66, 016104 (2002)

    2002

  28. [36]

    A. V. Goltsev, S. N. Dorogovtsev, and J. F. F. Mendes, Critical phenomena in networks, Physical Review E67, 026123 (2003)

    2003

  29. [37]

    Aurell, G

    E. Aurell, G. Del Ferraro, E. Dom ´ ınguez, and R. Mulet, Cavity master equation for the continuous time dynamics of discrete-spin models, Physical Review E95, 052119 (2017)

    2017

  30. [38]

    Dom ´ ınguez, D

    E. Dom ´ ınguez, D. Machado, and R. Mulet, The cavity master equation: average and fixed point of the ferromagnetic model in random graphs, Journal of Statistical Mechanics: Theory and Experiment2020, 073304 (2020)

    2020

  31. [39]

    Aurell, D

    E. Aurell, D. Machado Perez, and R. Mulet, A closure for the master equation starting from the dynamic cavity method, Journal of Physics A: Mathematical and Theoretical56, 17LT02 (2023)

    2023

  32. [40]

    Ortega, D

    E. Ortega, D. Machado, and A. Lage-Castellanos, Dynamics of epidemics from cavity master equations: Susceptible- infectious-susceptible models, Physical Review E105, 024308 (2022)

    2022

  33. [41]

    F. L. Metz, Dynamical mean-field theory of complex systems on sparse directed networks, Physical Review Letters134, 037401 (2025)

    2025

  34. [42]

    Aguilera, S

    M. Aguilera, S. A. Moosavi, and H. Shimazaki, A unifying framework for mean-field theories of asymmetric kinetic ising systems, Nature communications12, 1197 (2021)

    2021

  35. [43]

    Aguilera, M

    M. Aguilera, M. Igarashi, and H. Shimazaki, Nonequilibrium thermodynamics of the asymmetric sherrington-kirkpatrick model, Nature Communications14, 3685 (2023)

    2023

  36. [44]

    Aguilera, P

    M. Aguilera, P. A. Morales, F. E. Rosas, and H. Shimazaki, Explosive neural networks via higher-order interactions in curved statistical manifolds, Nature Communications16, 6511 (2025)

    2025

  37. [45]

    Barrio, S

    R. Barrio, S. Ib´ a˜ nez, and L. P´ erez, Homoclinic organization in the Hindmarsh–Rose model: A three parameter study, Chaos30, 053132 (2020)

    2020

  38. [46]

    Barrio, S

    R. Barrio, S. Ib´ a˜ nez, L. P´ erez, and S. Serrano, Spike-adding structure in fold/hom bursters, Communications in Nonlinear Science and Numerical Simulation83, 105100 (2020)

    2020

  39. [47]

    Barrio, S

    R. Barrio, S. Ib´ a˜ nez, L. P´ erez, and S. Serrano, Classification of fold/hom and fold/Hopf spike-adding phenomena, Chaos 31, 043120 (2021)

    2021

  40. [48]

    Serrano, M

    S. Serrano, M. A. Mart ´ ınez, and R. Barrio, Order in chaos: Structure of chaotic invariant sets of square-wave neuron models, Chaos31, 043108 (2021)

    2021

  41. [49]

    Barrio, S

    R. Barrio, S. Ib´ a˜ nez, and L. P´ erez, Exploring the geometry of the bifurcation sets in parameter space, Scientific Reports 14, 10900 (2024)

    2024

  42. [50]

    A. J. Fontenele, N. A. P. de Vasconcelos, T. Feliciano, L. A. A. Aguiar, C. Soares-Cunha, B. Coimbra, L. Dalla Porta, S. Ribeiro, A. J. Rodrigues, N. Sousa, P. V. Carelli, and M. Copelli, Criticality between cortical states, Physical Review Letters122, 208101 (2019)

    2019

  43. [51]

    Copelli, Physics of psychophysics: Spikes matter for network-scaled Ornstein–Uhlenbeck models, Physical Review Research2, 013058 (2020)

    M. Copelli, Physics of psychophysics: Spikes matter for network-scaled Ornstein–Uhlenbeck models, Physical Review Research2, 013058 (2020)

    2020

  44. [52]

    Levina, J

    A. Levina, J. M. Herrmann, and T. Geisel, Dynamical synapses causing self-organized criticality in neural networks, Nature Physics3, 857 (2007)

    2007

  45. [53]

    Levina, J

    A. Levina, J. M. Herrmann, and T. Geisel, Phase transitions towards criticality in a neural system with adaptive interac- tions, Physical Review Letters102, 118110 (2009)

    2009

  46. [54]

    Safavi, M

    S. Safavi, M. Chalk, N. K. Logothetis, and A. Levina, Signatures of criticality in efficient coding networks, Proceedings of the National Academy of Sciences121, e2302730121 (2024)

    2024

  47. [55]

    Lombardi, H

    F. Lombardi, H. J. Herrmann, C. Perrone-Capano, D. Plenz, and L. de Arcangelis, Balance between excitation and inhibition controls the temporal organization of neuronal avalanches, Phys. Rev. Lett108, 228703 (2012)

    2012

  48. [56]

    Lombardi, H

    F. Lombardi, H. J. Herrmann, and L. de Arcangelis, Balance between excitation and inhibition determines 1/f power spectrum in neuronal networks, Chaos27, 047402 (2017)

    2017

  49. [57]

    de Arcangelis, C

    L. de Arcangelis, C. Perrone-Capano, and H. Herrmann, Self-organized criticality model for brain plasticity, Phys. Rev. Lett96, 028107 (2006)

    2006

  50. [58]

    Guislain and E

    L. Guislain and E. Bertin, Nonequilibrium phase transition to temporal oscillations in mean-field spin models, Physical Review Letters130, 207102 (2023)

    2023

  51. [59]

    Guislain and E

    L. Guislain and E. Bertin, Discontinuous phase transition from ferromagnetic to oscillating states in a nonequilibrium mean-field spin model, Physical Review E109, 034131 (2024)

    2024

  52. [60]

    Guislain and E

    L. Guislain and E. Bertin, Tailoring the overlap distribution in driven mean-field spin models, Physical Review B109, 15 184203 (2024)

    2024

  53. [61]

    Guislain and E

    L. Guislain and E. Bertin, Collective oscillations in a three-dimensional spin model with non-reciprocal interactions, Journal of Statistical Mechanics: Theory and Experiment2024, 093210 (2024)

    2024

  54. [62]

    Guislain and E

    L. Guislain and E. Bertin, Hidden collective oscillations in a disordered mean-field spin model with non-reciprocal inter- actions, Journal of Physics A: Mathematical and Theoretical57, 375001 (2024)

    2024

  55. [63]

    Guislain and E

    L. Guislain and E. Bertin, Far-from-equilibrium complex landscapes, Physical Review E111, L062101 (2025)

    2025

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.