arxiv: 2605.00729 · v1 · submitted 2026-05-01 · 🧮 math.ST · math.PR· nlin.AO· stat.OT· stat.TH

Recognition: unknown

Intermittency induced by long memory under stochastic regime switching

Mauricio Herrera-Mar\'in

Pith reviewed 2026-05-09 18:27 UTC · model grok-4.3

classification 🧮 math.ST math.PRnlin.AOstat.OTstat.TH

keywords intermittencylong memoryregime switchingVolterra equationsannealed-quenched dichotomyself-exciting processesstochastic stability

0 comments

The pith

Long memory combined with stochastic regime switching induces intermittency where systems stable in expectation exhibit pathwise growth and burst phases.

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

The paper examines nonlinear dynamical systems governed by Volterra integral equations whose memory kernels and excitation operators are modulated by an ergodic Markov chain. It shows that under an averaged subcriticality condition, the system remains stable in mean-square sense with uniform moment bounds derived from a memory-adapted Lyapunov functional. However, along individual sample paths, rare excursions into supercritical regimes are amplified by the long memory, leading to intermittent macroscopic bursts with heavy tails and a deterministic almost-sure growth rate. This annealed-quenched separation is specific to non-Markovian systems and is transferred from microscopic point process models to the macroscopic limit. Numerical illustrations confirm that burst localization depends on the underlying graph and noncommuting operators.

Core claim

In network-coupled operator-valued Volterra evolutions with completely monotone memory kernels modulated by an ergodic continuous-time Markov chain, an averaged memory gain yields annealed stability in expectation, while quenched behavior features almost sure growth and metastable burst phases amplified by memory during persistent supercritical regimes.

What carries the argument

The annealed-quenched dichotomy arising from the interaction of long-range memory kernels and stochastic regime switching, formalized via memory-adapted Lyapunov functionals for the annealed side and subadditive ergodic arguments for the quenched growth exponent.

If this is right

Uniform moment bounds and mean-square control hold under the averaged subcriticality condition.
Pathwise almost sure growth occurs with a deterministic exponent obtained via subadditive ergodic theory.
Microscopic regime-modulated self-exciting point processes converge to the random-coefficient Volterra macroscopic limit.
Intermittent bursts exhibit heavy-tailed statistics and depend on graph geometry and noncommuting excitation operators.

Where Pith is reading between the lines

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

Real-world systems with long memory and occasional parameter shifts may exhibit unexpected intermittency not captured by Markovian models alone.
The micro-macro correspondence allows branching process simulations to predict burst statistics in the continuous Volterra setting.
Relaxing ergodicity of the switching chain could produce qualitatively different growth behaviors.
Network topology choices may offer a way to suppress or localize bursts through geometry effects.

Load-bearing premise

The memory kernels are completely monotone and the regime-switching Markov chain is ergodic with the excitation operators satisfying the stated subcriticality condition.

What would settle it

A numerical simulation or analytical counterexample where completely monotone kernels are replaced by non-monotone ones while keeping other conditions, showing loss of uniform moment bounds and separation of annealed and quenched behaviors.

Figures

Figures reproduced from arXiv: 2605.00729 by Mauricio Herrera-Mar\'in.

**Figure 1.** Figure 1: Experiment I: burst amplification under two-regime switching. (A) Representative sample path of ∥U(t)∥X with unstable regime intervals shaded. (B) Modewise projections onto dominant Laplacian eigenmodes, showing that burst amplification is spectrally concentrated. (C) Empirical CCDF of burst sizes B, exhibiting an approximate power-law region consistent with Theorem 6.5. trajectory ∥U(t)∥X with regime in… view at source ↗

**Figure 2.** Figure 2: Experiment II: memory-driven intermittency. Increasing α (stronger memory) and decreasing θ (weaker tempering) broaden burst tails and shift the distribution of γT toward positive values, even when annealed summaries remain moderate on [0, T] view at source ↗

**Figure 3.** Figure 3: Experiment III: switching-rate phase diagram (annealed vs. quenched signatures). The intermittent band appears where annealed summaries remain controlled while Pburst is nontrivial and Med[γT ] > 0, consistent with Proposition 6.7. produce heavier tails. Mechanistically, long residence times in U provide the exposure while long memory increases the accumulated unstable feedback. 8.6 Experiment III: Switch… view at source ↗

**Figure 4.** Figure 4: Experiment IV: noncommutative excitation operators. Noncommutativity induces mode mixing (bursts no longer align with a single Laplacian mode), yet burst directions remain concentrated in a low-dimensional subspace and burst tails remain heavy. the residence-time mechanism in Theorem 6.5. 8.7 Experiment IV: Noncommutative excitation operators and mode mixing To stress-test modal intuition, we construct ex… view at source ↗

**Figure 5.** Figure 5: Experiment V: network size and geometry. Topology primarily modulates where bursts concentrate in node space (IPR scaling), while dominant spectral activation typically occurs in intermediate Laplacian bands. rather than by a single slow mode. Localization exhibits a strong and geometry-dependent size law: IPR(z ∗ ) decreases with n but with systematic offsets across topologies, reflecting differences in e… view at source ↗

**Figure 6.** Figure 6: Experiment VI: numerical micro–macro validation under switching. The relative discrepancy Errrel N (T) decreases systematically with N, consistent with lawof-large-numbers scaling, even along environment paths that generate strong near-critical bursts. remain controlled on operational horizons while quenched diagnostics show bursty, heavytailed amplification driven by long residence times and long memory… view at source ↗

read the original abstract

We study a fundamental instability mechanism in nonlinear, nonlocal dynamical systems arising from the interaction of long-range memory and stochastic regime switching. The dynamics are governed by network-coupled, operator-valued Volterra evolutions with completely monotone memory kernels whose excitation operators and kernel parameters are modulated by an ergodic finite-state continuous-time Markov chain. We formalize a sharp separation between annealed stability (in expectation) and quenched behaviour (along typical sample paths). On the annealed side, we identify an averaged memory gain that yields uniform moment bounds and a memory-adapted Lyapunov functional implying mean-square control under an averaged subcriticality condition. On the quenched side, we show that rare but persistent excursions into supercritical regimes are amplified by memory, producing intermittent macroscopic bursts with heavy-tailed statistics and a deterministic almost sure growth exponent obtained via a subadditive ergodic argument. This establishes an annealed--quenched dichotomy specific to non-Markovian switching systems, where stability in expectation can coexist with pathwise growth and metastable burst phases. We further derive a micro--macro correspondence by proving that a population of regime-modulated self-exciting point processes converges, both annealed and quenched, to the random-coefficient Volterra limit, transferring the burst mechanism from microscopic branching dynamics to macroscopic long-memory flows. Numerical experiments illustrate how burst localization depends on graph geometry and on noncommuting excitation operators.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The annealed-quenched split via memory-amplified bursts and the point-process convergence are the actual novelties, but the deterministic a.s. growth on the quenched side rests on a subadditive argument that long memory likely undermines.

read the letter

The paper's main claim is that long memory plus ergodic regime switching produces stability in expectation while allowing pathwise intermittent growth and metastable bursts in a Volterra system. They also show a population of modulated self-exciting point processes converges to the macroscopic random-coefficient limit, carrying the burst statistics along both annealed and quenched senses.

Referee Report

1 major / 3 minor

Summary. The paper studies intermittency in nonlinear nonlocal systems from the interaction of long-range memory and stochastic regime switching. It considers network-coupled operator-valued Volterra evolutions with completely monotone kernels modulated by an ergodic finite-state continuous-time Markov chain. The central claims are an annealed-quenched dichotomy: annealed mean-square stability under an averaged subcriticality condition via memory-adapted Lyapunov functionals and uniform moment bounds, versus quenched pathwise intermittent macroscopic bursts with heavy-tailed statistics and a deterministic a.s. growth exponent via subadditive ergodic arguments. It further establishes a micro-macro correspondence by showing convergence of regime-modulated self-exciting point processes to the random-coefficient Volterra limit (annealed and quenched), with numerics on burst localization depending on graph geometry and noncommuting operators.

Significance. If the results hold, this identifies a specific mechanism for intermittency in non-Markovian switching systems where expectation stability coexists with pathwise growth and metastable bursts, extending standard tools (Lyapunov functionals, subadditive ergodic theorem, point-process convergence) to this setting. The micro-macro correspondence transferring burst dynamics from microscopic branching to macroscopic long-memory flows is a clear strength, as is the explicit separation of annealed and quenched behaviors. This could inform analysis of complex systems with memory and modulation, provided the quenched arguments are fully justified.

major comments (1)

[Quenched analysis (subadditive ergodic argument)] Quenched analysis, subadditive ergodic argument for the deterministic a.s. growth exponent: the regime-modulated Volterra process has state depending on the full history through the completely monotone kernel. Standard subadditive ergodic theorems (e.g., Kingman-type) require subadditivity on an ergodic dynamical system with suitable measurability and mixing; the long-range dependence from the memory tail may destroy independence or mixing of increments even when the modulating Markov chain is ergodic. The manuscript must supply explicit control on the memory-induced dependence to ensure the limit is deterministic and pathwise constant, as this is load-bearing for the quenched side of the annealed-quenched dichotomy.

minor comments (3)

[Notation and setup] The notation for excitation operators, kernel parameters, and the subcriticality condition should be introduced with a brief example of noncommuting operators to aid readability before the main theorems.
[Numerical experiments] Numerical experiments section: figure captions for burst localization should explicitly list the graph geometries, parameter values, and how noncommuting operators are implemented to allow reproduction.
[References] References: add precise citations for the specific statements of the subadditive ergodic theorem and point-process convergence results used, including any extensions to Volterra processes.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and for isolating the key technical point in the quenched analysis. We address it directly below and will revise the manuscript to make the required control explicit.

read point-by-point responses

Referee: Quenched analysis, subadditive ergodic argument for the deterministic a.s. growth exponent: the regime-modulated Volterra process has state depending on the full history through the completely monotone kernel. Standard subadditive ergodic theorems (e.g., Kingman-type) require subadditivity on an ergodic dynamical system with suitable measurability and mixing; the long-range dependence from the memory tail may destroy independence or mixing of increments even when the modulating Markov chain is ergodic. The manuscript must supply explicit control on the memory-induced dependence to ensure the limit is deterministic and pathwise constant, as this is load-bearing for the quenched side of the annealed-quenched dichotomy.

Authors: We agree that the long memory induced by the completely monotone kernel requires explicit justification before Kingman’s theorem can be invoked. Because the kernels are completely monotone, Bernstein’s theorem supplies a representation as Laplace transforms of positive measures. This permits an auxiliary-state augmentation (a continuum of exponential modes) that renders the joint (Volterra + auxiliary + Markov chain) process Markovian with respect to a suitable filtration. The finite-state ergodicity of the modulating chain then propagates to the extended system, restoring the necessary mixing and allowing the subadditive sequence (log-norm of the solution) to satisfy the hypotheses of the subadditive ergodic theorem. The resulting limit is therefore deterministic and pathwise constant. We will add a dedicated appendix that (i) constructs the auxiliary process, (ii) verifies the required measurability and subadditivity, and (iii) confirms that the ergodic theorem applies directly to the augmented dynamics. This addresses the load-bearing character of the quenched side of the dichotomy. revision: yes

Circularity Check

0 steps flagged

No circularity: claims rely on external ergodic theory and Lyapunov methods

full rationale

The derivation establishes an annealed-quenched dichotomy for non-Markovian regime-switching Volterra systems using an averaged memory gain for uniform moment bounds and a memory-adapted Lyapunov functional under subcriticality, together with a subadditive ergodic argument for the deterministic almost-sure growth on the quenched side. These steps invoke standard external tools (ergodic theorems, Lyapunov analysis) applied to the modulated kernels and Markov chain, without any reduction of a claimed prediction or growth exponent to a fitted parameter, self-defined quantity, or self-cited uniqueness result by construction. The micro-macro convergence from point processes to the random-coefficient Volterra limit is likewise obtained via independent convergence arguments. No load-bearing step equates outputs to inputs via renaming, ansatz smuggling, or self-referential fitting.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on standard domain assumptions from stochastic analysis rather than new free parameters or invented entities.

axioms (2)

domain assumption Memory kernels are completely monotone.
Invoked to guarantee positivity and the existence of the memory-adapted Lyapunov functional for annealed bounds.
domain assumption The modulating continuous-time Markov chain is ergodic.
Used to obtain the averaged memory gain and to apply ergodic theorems for the quenched growth exponent.

pith-pipeline@v0.9.0 · 5543 in / 1349 out tokens · 57510 ms · 2026-05-09T18:27:41.535950+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

26 extracted references

[1]

Uriel Frisch.Turbulence: The Legacy of A. N. Kolmogorov. Cambridge University Press, Cambridge, 1995

1995
[2]

Universal scaling laws in fully developed tur- bulence.Physical Review Letters, 72(3):336–339, 1994

Zhen-Su She and Emmanuel Leveque. Universal scaling laws in fully developed tur- bulence.Physical Review Letters, 72(3):336–339, 1994

1994
[3]

Intermittent transition to turbulence in dissipa- tive dynamical systems.Communications in Mathematical Physics, 74(2):189–197, 1980

Yves Pomeau and Paul Manneville. Intermittent transition to turbulence in dissipa- tive dynamical systems.Communications in Mathematical Physics, 74(2):189–197, 1980

1980
[4]

Platt, E

N. Platt, E. A. Spiegel, and C. Tresser. On–off intermittency: A mechanism for bursting.Physical Review Letters, 70(3):279–282, 1993

1993
[5]

Beggs and Dietmar Plenz

John M. Beggs and Dietmar Plenz. Neuronal avalanches in neocortical circuits. Journal of Neuroscience, 23(35):11167–11177, 2003

2003
[6]

Empirical properties of asset returns: stylized facts and statistical issues

Rama Cont. Empirical properties of asset returns: stylized facts and statistical issues. Quantitative Finance, 1(2):223–236, 2001

2001
[7]

Birkhäuser, 1993

Jan Prüss.Evolutionary integral equations and applications. Birkhäuser, 1993

1993
[8]

PhD thesis, Eindhoven University of Technology, 2000

Elena Bazhlekova.Fractional evolution equations in Banach spaces. PhD thesis, Eindhoven University of Technology, 2000

2000
[9]

Horwood, 1999

Xuerong Mao.Stochastic Differential Equations and Applications. Horwood, 1999. 31 Intermittency induced by long memory under stochastic regime switchingarXiv preprintmath.DS

1999
[10]

Yin and C

G. Yin and C. Zhu.Hybrid Switching Diffusions. Springer, 2010

2010
[11]

Keeling and Pejman Rohani.Modeling Infectious Diseases in Humans and Animals

Matt J. Keeling and Pejman Rohani.Modeling Infectious Diseases in Humans and Animals. Princeton University Press, 2008

2008
[12]

Epidemic processes in complex networks.Reviews of Modern Physics, 87(3):925–979, 2015

Romualdo Pastor-Satorras, Claudio Castellano, Piet Van Mieghem, and Alessandro Vespignani. Epidemic processes in complex networks.Reviews of Modern Physics, 87(3):925–979, 2015

2015
[13]

Cambridge University Press, 2014

Wulfram Gerstner, Werner Kistler, Richard Naud, and Liam Paninski.Neuronal Dynamics. Cambridge University Press, 2014

2014
[14]

Eden, Matthew R

Wilson Truccolo, Uri T. Eden, Matthew R. Fellows, John P. Donoghue, and Emery N. Brown. A point process framework for relating neural spiking activity to spiking history.Journal of Neurophysiology, 93(2):1074–1089, 2005

2005
[15]

Robert F. Engle. Autoregressive conditional heteroskedasticity with estimates of the variance of uk inflation.Econometrica, 50(4):987–1007, 1982

1982
[16]

Hawkes processes in finance.Market Microstructure and Liquidity, 1(01):1550005, 2015

Emmanuel Bacry, Iacopo Mastromatteo, and Jean-François Muzy. Hawkes processes in finance.Market Microstructure and Liquidity, 1(01):1550005, 2015

2015
[17]

Franzke, Terence J

Christian L.E. Franzke, Terence J. O’Kane, Judith Berner, Paul D. Williams, and Valerio Lucarini. Stochastic climate theory and modeling.WIREs Climate Change, 6(1):63–78, 2015

2015
[18]

Cambridge University Press, 2013

Shaun Lovejoy and Daniel Schertzer.The Weather and Climate: Emergent Laws and Multifractal Cascades. Cambridge University Press, 2013

2013
[19]

Spectra of some self-exciting and mutually exciting point processes

Alan G Hawkes. Spectra of some self-exciting and mutually exciting point processes. Biometrika, 58(1):83–90, 1971

1971
[20]

Stability of nonlinear hawkes processes

Pierre Brémaud and Laurent Massoulié. Stability of nonlinear hawkes processes. Annals of Probability, 24(3):1563–1588, 1996

1996
[21]

Limit theorems for nearly unstable hawkes processes.Annals of Applied Probability, 25(2):600–631, 2015

Thibault Jaisson and Mathieu Rosenbaum. Limit theorems for nearly unstable hawkes processes.Annals of Applied Probability, 25(2):600–631, 2015

2015
[22]

Rough fractional diffusions as scaling limits of nearly unstable hawkes processes.Annals of Applied Probability, 26(5):3540– 3570, 2016

Thibault Jaisson and Mathieu Rosenbaum. Rough fractional diffusions as scaling limits of nearly unstable hawkes processes.Annals of Applied Probability, 26(5):3540– 3570, 2016

2016
[23]

Chevallier

J. Chevallier. Mean-field limit of generalized hawkes processes.Stochastic Processes and their Applications, 129:420–435, 2019

2019
[24]

L. N. Trefethen, A. E. Trefethen, S. C. Reddy, and T. A. Driscoll. Hydrodynamic stability without eigenvalues.Science, 261:578–584, 1993. 32 Intermittency induced by long memory under stochastic regime switchingarXiv preprintmath.DS

1993
[25]

Benaych-Georges and R

F. Benaych-Georges and R. R. Nadakuditi. The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices.Advances in Mathematics, 227:494–521, 2011

2011
[26]

Bordenave and M

C. Bordenave and M. Lelarge. Resolvent of large random graphs.Random Structures & Algorithms, 37:332–352, 2010. 33

2010