Reservoir Computing based on Quenched Chaos

Jaesung Choi; Pilwon Kim

REVIEW 4 major objections 7 minor 1 cited by

Reservoir Computing based on Quenched Chaos

T0 review · 4 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A reservoir made of coupled chaotic Lorenz oscillators, held at the explosive-death transition, achieves test errors at least 1000 times smaller than a reservoir of regular phase oscillators at explosive synchronization.

desk verdict A plausible and internally controlled reservoir-computing result whose 1000x superiority claim is a single-run observation resting on an untested basin-stability assumption. read the letter →

arxiv 1909.01571 v1 pith:6LR7J4DG submitted 2019-09-04 nlin.CD cs.CCcs.LGcs.NE

Jaesung Choi , Pilwon Kim This is my paper · ORCID

classification nlin.CDcs.CCcs.LGcs.NE MSC 37D4534C15 PACS 05.45.-a05.45.Xt

keywords reservoircomputingquenchedchaosexplosivedeathamplitudecoupledLorenzoscillatorscriticalityinformationcapacitysynchronization

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 tries to show that a reservoir computer built from chaotic Lorenz oscillators works best when the oscillators sit exactly at the explosive-death transition, the coupling strength at which their oscillations abruptly collapse into a quiet equilibrium. At this critical point, the system offers a rich variety of transient orbits for computation while still returning to a stable ground state after each input. The authors report that this 'quenched chaos' reservoir has test errors at least 1000 times smaller than a reservoir of regular phase oscillators poised at explosive synchronization, across three signal-reconstruction tasks. They also report that total information capacity, computed without training, predicts which reservoir performs better. If true, the paper establishes that chaos, properly quenched at a first-order transition, is a useful building block for critical reservoirs.

What carries the argument

The engine is the coupled Lorenz system in Eq. (1), N nonidentical Lorenz oscillators with mean-field diffusion $K(Q\bar{x}-x_i)$, natural frequencies uniformly distributed in $[1,1.3]$, and $Q=0.7$. Backward continuation in Figure 1 yields an order-parameter phase diagram showing a discontinuous drop in $r_{var}$ and $A$ along a diagonal line in the $(\rho,K)$ plane, and the reservoir is set near that line. The readout, Eq. (2), is a linear map on the last ten sampled values of each frequency derivative $x_i'$, trained by least squares. Explosive death supplies both ingredients: a stable amplitude-death ground state that erases previous inputs, and a nearby chaotic phase whose high-dimensional transient orbits give the readout a large repertoire.

What would settle it

Integrate Eq. (1) forward from random initial conditions for $\rho=28$, $Q=0.7$, near $K=2.2$ and measure $A(K)=\langle x_{i,\max}\rangle-\langle x_{i,\min}\rangle$: if the oscillation amplitude declines continuously with $K$, or if an infinitesimally perturbed death state spontaneously resumes oscillations for the same $K$ used for computing, then the transition is not explosive and the ground state is not attracting as claimed. Independently, rerun the three tasks at $N=150$ with the $(10,0.1)$ readout; if the normalized test errors of the quenched-chaos reservoir are not at least three orders of magnitude below the explosive-synchronization reservoir on all three tasks, the headline quantitative claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a network of nonidentical Lorenz oscillators coupled by mean-field diffusion, Eq. (1), undergoes explosive death, a discontinuous, first-order transition from oscillatory motion to complete amplitude death, and that operating the network exactly at that transition creates a 'quenched chaos' reservoir. The reservoir's ground state is the dead equilibrium; input perturbations excite transient orbits that retain chaotic richness, and the readout maps ten past sampled frequency derivatives to the target. The paper reports that this reservoir outperforms the explosive-synchronization reservoir by at least a factor of 1000 in normalized test error on three tasks, inferring hidden variables of the Rössler system and Chua's circuit, and filtering a Mackey-Glass input. It also reports that the total information capacity measure matches this performance ordering, supporting the idea that capacity computed without training can predict computational capability at a critical point.

Load-bearing premise

The entire construction rests on the claim that the nonidentical coupled Lorenz system in Eq. (1) has a robust, discontinuous explosive-death transition at the chosen parameters, with a stable amplitude-death state that attracts the system back after each input; the paper supports this only with numerical backward continuation, not an analytic proof.

Editorial extensions

If this is right

Operating a continuous reservoir at the explosive-death transition makes chaos usable, because the dead state stabilizes the reservoir while nearby chaotic dynamics supply computational richness.
Total information capacity computed without task training can rank candidate critical reservoirs, so it can be used to choose operating parameters and network size before running expensive benchmarks.
Chaotic nodes outperform regular phase oscillators at the same kind of first-order criticality, suggesting that attractor dimension or the collapsed dimension of the reservoir is a relevant design axis.
The same recipe could produce a broad range of physical reservoirs in systems known to exhibit explosive death, such as lasers, chemical oscillators, or neuronal models.

Reading between the lines

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

A testable extension the paper does not run is to vary the frequency spread or $Q$ and check whether the 1000x margin tracks the size of the hysteresis loop of the explosive-death transition, not just its existence.
If information capacity is truly predictive, it should also predict relative performance within the quenched-chaos family, for example which of several chaotic oscillator types at their own explosive-death points gives the lowest error; this is an independent check of the paper's explanation.
The 'at least 1000 times smaller' figure is demonstrated for three benchmark tasks with one readout type; whether it survives across task families, input amplitudes, or noise is left open by the paper.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Short version: this is a conditional-accept-level paper. The central idea—using the explosive death transition of coupled Lorenz systems as a reservoir—is a genuine extension of your earlier ES reservoir work, and the authors earn credit for the strongest internal control in the paper: across three independent tasks, the error minima fall on exactly the same line in the (K, rho) plane as the explosive-death transition. That alignment is a real result, not a fitted one.

What's new: the specific combination of quenched chaotic oscillators at a first-order amplitude-death transition, with a comparison against explosive synchronization reservoirs, is not in the cited literature. The capacity comparison in Section 4.2 is a reasonable attempt to explain why QC beats ES, though it is post-hoc.

Soft spots, in order of softness. First, the framework leans on the assumption that the nonidentical Lorenz ensemble in Eq. (1) has a robust discontinuous death transition with an attracting death state, and that the reservoir returns to that state after each input perturbation. The authors demonstrate the transition only via backward continuation in Figure 1, and they never directly test basin stability under input. That is a real gap. However, the error maps in Figure 2 indirectly supply some evidence: low test error along the critical line means the reservoir is actually returning to a usable ground state across a range of K and rho. So the gap is not fatal, but it deserves a direct check—for example, perturbing from the death state at various amplitudes and measuring return times.

Second, there are no error bars anywhere. The 1000x claim is a single numerical run per configuration, and the reader cannot tell whether it is robust to initial conditions, integration tolerances, or readout details. No code or data is released, so reproducing these numbers means reconstructing the setup from the text. This is the kind of thing that can sink an otherwise plausible result in this literature.

Third, the 'predictive measure' claim in the abstract overstates what is done. The total information capacity is computed after the fact and shown to agree with the task errors. That is a correlation, not a prediction. The paper should say so.

The citation pattern is fine—the only self-citation is the ES baseline [7], which is exactly where it belongs.

Who this is for: people working on physical reservoir computing or critical dynamics in coupled oscillators. It would be a reasonable journal or conference paper after revision. It deserves a serious referee, and I would send it to review rather than desk reject.

Referee Report

4 major / 7 minor

Summary. This paper proposes a reservoir computer based on coupled nonidentical Lorenz oscillators with mean-field diffusion, operated near the first-order explosive death (ED) transition, termed 'quenched chaos' (QC). The authors show that for three benchmark tasks (inferring missing variables of the Rössler and Chua systems, and filtering the Mackey-Glass time series), the test error is minimized along a line in the (ρ, K) plane that coincides with the ED critical line obtained by backward continuation. They then compare the QC reservoir with a reservoir of Kuramoto phase oscillators near explosive synchronization (ES) from their previous work, and report that QC errors are at least 1000 times smaller across all three tasks. They further compute the total information capacity from Dambre et al. for both reservoirs and find that the capacity curves predict the relative task performance.

Significance. If the claims hold, the paper provides a concrete demonstration that chaotic units, when kept near a first-order death transition, can serve as highly effective reservoirs, and that the information-capacity measure is a useful predictor of performance. The study's internal control—the alignment of error minima with the critical line in three independent tasks—is a strong point. The use of the external capacity measure from [9] reduces the risk of circularity in comparing the two reservoirs. The numerical protocol is clearly described, although some details are missing. The main weaknesses are the lack of repeated trials/error bars and the unverified stability of the death-state ground state under input perturbations.

major comments (4)

[Section 3, Figure 1] The claim that the system 'always returns to after every computation' (Section 3) is not supported by any numerical evidence for the nonidentical Lorenz system in Eq. (1). The ED transition is established only via backward continuation from large K, which follows the oscillatory branch; the forward basin of attraction of the amplitude-death state, and in particular its response to the input perturbations used in training, is not measured. Because the reservoir is operated near a first-order transition where bistability is expected, the reader cannot rule out that inputs occasionally drive the system onto the oscillatory branch, which would change the interpretation of the results. The authors should demonstrate the return to the ground state directly (e.g., by plotting the order parameter rvar or the state deviation over time after input pulses) and, ideally, quantify the basin stability for the input amplitudes used.
[Section 4.1, Figure 3] The central quantitative claim that QC errors are at least 1000 times smaller than ES errors is based on single runs. No error bars, standard deviations, or repeated trials with different network realizations (e.g., different random frequency sets or initial conditions) are reported. Since the reservoir is stochastic due to the uniform [1,1.3] frequency draws, the 1000x factor could be an artifact of a particular realization. The authors should provide means and spreads over multiple independent realizations, or at least state that the qualitative ordering is stable across realizations.
[Section 4.2] The information-capacity calculation is not described in sufficient detail for reproducibility. The paper states 'the most of the parameters for evaluation are from [9]' but does not specify how the Dambre et al. measure is applied to the continuous-time reservoirs, how many basis functions are used, the length of the input signals, or the discretization. Since the capacity comparison is used to support the claim of predictability, the calculation must be specified precisely.
[Sections 4.1-4.2] The comparison between QC and ES is potentially confounded by the difference in per-node dimension: a QC node is a three-variable Lorenz system whereas an ES node is a single phase variable. The authors themselves attribute the capacity difference to the larger collapsed attractor dimension, so the reported improvement may be due to higher-dimensional dynamics rather than to chaos per se. To substantiate the conclusion that 'chaotic nodes are more beneficial,' a control reservoir with nonchaotic but high-dimensional nodes would be needed.

minor comments (7)

[Section 1] There is a typo in the Introduction: 'In thiw work' should be 'In this work.'
[Section 2.1] The word 'rougly' appears in the description of computational cost; it should be 'roughly.'
[Section 4.1] The name 'Rssler' should be 'Rössler' throughout the paper.
[Section 4.2] In the phrase 'Hausedorﬀ dimension,' 'Hausedorﬀ' should be 'Hausdorff.'
[Section 2.2] The numerical integration scheme and time step used to solve Eqs. (1) and (7) are not stated; please provide them for reproducibility.
[Section 4.1] The exact operating parameters (K and ρ) on the critical line used for the QC-ES comparison in Figure 3 are not given; please state them explicitly.
[References] Reference [17] lists the conference year as 2019 in the journal name but the final year as 2017; please correct this inconsistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the quenched-chaos reservoir is evaluated on standard benchmark tasks with an independent capacity measure, and the self-cited ES baseline serves only as a comparison system.

full rationale

The paper's derivation chain is self-contained and does not reduce to its inputs. The explosive-death operating point is established numerically for the nonidentical Lorenz ensemble in Figure 1 via backward continuation, independent of task errors. The three tasks are standard benchmarks (Rössler inference, Chua inference, Mackey-Glass filtering), and the readout is a linear least-squares fit of past frequency samples, so the task results are genuine measurements of reservoir performance. The information capacity is computed from the externally developed measure of Dambre et al. [9]; its agreement with task errors is a post hoc correlation, not a fitted parameter renamed as a prediction. The only self-citation, [7], supplies the ES baseline reservoir for comparison; it is not an input to the QC dynamics or to the capacity calculation, and it does not force the claimed 1000x error ratio. The concern that the amplitude-death ground state's stability under input perturbations is not rigorously verified is a robustness or correctness issue, not circularity, because the paper does not define QC's success in terms of that assumption or fit the assumption to the outcome. No equation is used both as premise and conclusion, and no fitted value is presented as a prediction. Therefore the round-trip analysis finds no circular step.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new entities. Its central claim depends on standard Lorenz dynamics, a hand-chosen mean-field coupling, a hand-chosen frequency distribution, and a readout scheme, all stated explicitly. The main unproven ingredients are the existence and stability of the explosive death transition for nonidentical oscillators, and the validity of information capacity as a performance predictor.

free parameters (4)

Mean-field intensity Q = 0.7
Set by hand following reference [33]; controls the explosive death transition and is not varied or justified.
Natural frequency range = uniform in [1, 1.3]
Chosen by hand for the N=100 oscillators; affects the phase diagram and transition, no sensitivity analysis.
Readout type (s, Delta t) = (10, 0.1)
Fixed for all tasks; the past 10 sampled derivative values with step 0.1 define the readout feature space.
Number of nodes N = 100 (up to 200 in scaling tests)
Used in the phase diagram and tasks; could affect the criticality and performance, though trends are shown up to 200.

assumptions (4)

domain assumption The explosive death transition found in identical oscillators by Verma et al. [33] extends to nonidentical oscillators with frequencies in [1, 1.3].
Equation (1) generalizes the mean-field coupling of [33] to nonidentical wi, and Figure 1 shows a discontinuous transition numerically, but no analytic proof is given.
domain assumption The amplitude death state is a stable ground state that the reservoir returns to after each input.
Section 3 states this property ('the system always returns to after every computation'), but stability is not proven; it is assumed for the readout scheme.
domain assumption The total information capacity from Dambre et al. [9] is a valid measure of a reservoir's computational capability.
Section 4.2 uses this external measure; the paper does not derive it and assumes it reflects task performance.
domain assumption The readout based solely on past values of x'_i (time-derivatives of x) provides sufficient features for the tasks.
The (s, Delta t)-type readout in Eq. (2) is chosen without comparison to other state variables or readout types.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Reservoir Computing based on Quenched Chaos." pith.science (2026). https://pith.science/paper/6LR7J4DG

@misc{pith2026190901571,
  author       = {Pith},
  title        = {Pith review of: Reservoir Computing based on Quenched Chaos},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6LR7J4DG}},
  note         = {Machine review of arXiv:1909.01571}
}

read the original abstract

Reservoir computing(RC) is a brain-inspired computing framework that employs a transient dynamical system whose reaction to an input signal is transformed to a target output. One of the central problems in RC is to find a reliable reservoir with a large criticality, since computing performance of a reservoir is maximized near the phase transition. In this work, we propose a continuous reservoir that utilizes transient dynamics of coupled chaotic oscillators in a critical regime where sudden amplitude death occurs. This "explosive death" not only brings the system a large criticality which provides a variety of orbits for computing, but also stabilizes them which otherwise diverge soon in chaotic units. The proposed framework shows better results in tasks for signal reconstructions than RC based on explosive synchronization of regular phase oscillators. We also show that the information capacity of the reservoirs can be used as a predictive measure for computational capability of a reservoir at a critical point.

Figures

Figures reproduced from arXiv: 1909.01571 by the authors.

**Figure 2.** Test errors according to the coupling K and lorenz system parameter ρ. (a) inferring a missing variable of the Rssler system. (b) inferring a missing variable of the Chua’s circuit. (c) filtering the Mackey-Glass equation assures that the computational performance of the reservoirs is maximized near the first order phase transition. 4.2 Information capacity of regular and chaotic reservoirs In the previous work [7],… view at source ↗

**Figure 3.** Comparison of test errors of QC and ES with respect to nu [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Total information capacity of QC and ES with respect to nu [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Transient Dynamics in Lattices of Differentiating Ring Oscillators
cs.NE 2025-06 conditional novelty 6.0 of 10

Large lattices of differentiating ring oscillators form growing locally synchronized domains whose steady-state scale depends on the lattice coupling geometry.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages · cited by 1 Pith paper

[7]

Critical neuromorphic computing bas ed on explo- sive synchronization

Jaesung Choi and Pilwon Kim. Critical neuromorphic computing bas ed on explo- sive synchronization. Chaos: An Interdisciplinary Journal of Nonlinear Science , 29(4):043110, 2019

work page 2019
[9]

Informa- tion processing capacity of dynamical systems

Joni Dambre, David Verstraeten, Benjamin Schrauwen, and Se rge Massar. Informa- tion processing capacity of dynamical systems. Scientiﬁc reports , 2:514, 2012

work page 2012
[1]

Chaotic oscillations and bifurcations in squid gian t axons

Kazuyuki Aihara. Chaotic oscillations and bifurcations in squid gian t axons. Chaos, pages 257–269, 1986

work page 1986
[2]

Computation with cha os: A paradigm for cortical activity

Agnesa Babloyantz and Carlos Louren¸ co. Computation with cha os: A paradigm for cortical activity. Proceedings of the National Academy of Sciences , 91(19):9027–9031, 1994

work page 1994
[3]

Neuronal avalanches in neocort ical circuits

John M Beggs and Dietmar Plenz. Neuronal avalanches in neocort ical circuits. Jour- nal of neuroscience , 23(35):11167–11177, 2003

work page 2003
[4]

Being critical of criticality in the br ain

John M Beggs and Nicholas Timme. Being critical of criticality in the br ain. Frontiers in physiology , 3:163, 2012

work page 2012
[5]

Marker s of criticality in phase synchronization

Maria Botcharova, Simon F Farmer, and Luc Berthouze. Marker s of criticality in phase synchronization. Frontiers in systems neuroscience , 8:176, 2014

work page 2014
[6]

Nonlinear prediction of chaotic time series

Martin Casdagli. Nonlinear prediction of chaotic time series. Physica D: Nonlinear Phenomena, 35(3):335–356, 1989. 8

work page 1989

Show all 37 references

[8]

Experimental and theore tical studies of a coupled chemical oscillator: phase death, multistability and in-phas e and out-of- phase entrainment

Michael F Crowley and Irving R Epstein. Experimental and theore tical studies of a coupled chemical oscillator: phase death, multistability and in-phas e and out-of- phase entrainment. The Journal of Physical Chemistry , 93(6):2496–2502, 1989

1989
[10]

Criticalit y meets learn- ing: Criticality signatures in a self-organizing recurrent neural net work

Bruno Del Papa, Viola Priesemann, and Jochen Triesch. Criticalit y meets learn- ing: Criticality signatures in a self-organizing recurrent neural net work. PloS one , 12(5):e0178683, 2017

2017
[11]

Coupled chaotic chemical oscillato rs

Milos Dolnik and Irving R Epstein. Coupled chaotic chemical oscillato rs. Physical Review E, 54(4):3361, 1996

1996
[12]

Reservoir computing using dynamic memristors for temporal inform ation processing

Chao Du, Fuxi Cai, Mohammed A Zidan, Wen Ma, Seung Hwan Lee, a nd Wei D Lu. Reservoir computing using dynamic memristors for temporal inform ation processing. Nature communications, 8(1):2204, 2017

2017
[13]

Oscillator death in systems of couple d neural oscilla- tors

GB Ermentrout and N Kopell. Oscillator death in systems of couple d neural oscilla- tors. SIAM Journal on Applied Mathematics , 50(1):125–146, 1990

1990
[14]

Reservoir computing : Quo vadis? In Proceedings of the 3rd ACM International Conference on Nano scale Computing and Communication, page 13

Alireza Goudarzi and Christof Teuscher. Reservoir computing : Quo vadis? In Proceedings of the 3rd ACM International Conference on Nano scale Computing and Communication, page 13. ACM, 2016

2016
[15]

Experimen tal observation of the amplitude death eﬀect in two coupled nonlinear oscillators

Ramon Herrero, M Figueras, J Rius, F Pi, and G Orriols. Experimen tal observation of the amplitude death eﬀect in two coupled nonlinear oscillators. Physical review letters, 84(23):5312, 2000

2000
[16]

Harnessing nonlinearity: Pre dicting chaotic systems and saving energy in wireless communication

Herbert Jaeger and Harald Haas. Harnessing nonlinearity: Pre dicting chaotic systems and saving energy in wireless communication. science, 304(5667):78–80, 2004

2004
[17]

Jensen and Gunnar Tufte

Johannes H. Jensen and Gunnar Tufte. Reservoir computing w ith a chaotic circuit. The 2019 Conference on Artiﬁcial Life , (29):222–229, 2017

2019
[18]

Computation at the edge of chaos: phase tra nsitions and emergent computation

Chris G Langton. Computation at the edge of chaos: phase tra nsitions and emergent computation. Physica D: Nonlinear Phenomena , 42(1-3):12–37, 1990

1990
[19]

Edge of chaos and pre diction of computa- tional performance for neural circuit models

Robert Legenstein and Wolfgang Maass. Edge of chaos and pre diction of computa- tional performance for neural circuit models. Neural Networks , 20(3):323–334, 2007

2007
[20]

Explosive ﬁrst-order transition to synchrony in networked chaotic oscillators

I Leyva, R Sevilla-Escoboza, JM Buld´ u, I Sendina-Nadal, J G´ omez-Garde˜ nes, A Are- nas, Y Moreno, S G´ omez, R Jaimes-Re´ ategui, and S Boccaletti. Explosive ﬁrst-order transition to synchrony in networked chaotic oscillators. Physical review letters , 108(16):168702, 2012

2012
[21]

Attention-locked computation with chaotic neural nets

Carlos Louren¸ co. Attention-locked computation with chaotic neural nets. Interna- tional Journal of Bifurcation and Chaos , 14(02):737–760, 2004. 9

2004
[22]

Dynamical reservoir properties as networ k eﬀects

Carlos Louren¸ co. Dynamical reservoir properties as networ k eﬀects. In ESANN, pages 503–508, 2006

2006
[23]

Dynamical computation reservoir emerging within a biological model network

Carlos Louren¸ co. Dynamical computation reservoir emerging within a biological model network. Neurocomputing, 70(7-9):1177–1185, 2007

2007
[24]

Structured reservoir computing with spat iotemporal chaotic at- tractors

Carlos Louren¸ co. Structured reservoir computing with spat iotemporal chaotic at- tractors. In ESANN, pages 501–506, 2007

2007
[25]

Evolution in materio: Looking beyon d the silicon box

Julian F Miller and Keith Downing. Evolution in materio: Looking beyon d the silicon box. In Proceedings 2002 NASA/DoD Conference on Evolvable Hardwar e, pages 167–176. IEEE, 2002

2002
[26]

Colloquium: Criticality and dynamical scaling in living s ystems

Miguel A Munoz. Colloquium: Criticality and dynamical scaling in living s ystems. Reviews of Modern Physics , 90(3):031001, 2018

2018
[27]

Strong coupling of nonlinear electronic and biological oscillators: reaching th e amplitude death regime

I Ozden, S Venkataramani, MA Long, BW Connors, and A V Nurmik ko. Strong coupling of nonlinear electronic and biological oscillators: reaching th e amplitude death regime. Physical review letters , 93(15):158102, 2004

2004
[28]

Amplitude dea th in complex networks induced by environment

V Resmi, G Ambika, RE Amritkar, and G Rangarajan. Amplitude dea th in complex networks induced by environment. Physical Review E , 85(4):046211, 2012

2012
[29]

Amplitude death with mean-ﬁe ld diﬀusion

Amit Sharma and Manish Dev Shrimali. Amplitude death with mean-ﬁe ld diﬀusion. Physical Review E , 85(5):057204, 2012

2012
[30]

Dynamics based computation

Sudeshna Sinha and William L Ditto. Dynamics based computation. physical review Letters, 81(10):2156, 1998

1998
[31]

Computing with distributed chao s

Sudeshna Sinha and William L Ditto. Computing with distributed chao s. Physical Review E, 60(1):363, 1999

1999
[32]

Recent advances in physical reservoir computing: a review

Gouhei Tanaka, Toshiyuki Yamane, Jean Benoit H´ eroux, Ryo sho Nakane, Naoki Kanazawa, Seiji Takeda, Hidetoshi Numata, Daiju Nakano, and Ak ira Hirose. Recent advances in physical reservoir computing: a review. Neural Networks , 2019

2019
[33]

Explosive death induced by mean–ﬁeld diﬀusion in ide ntical oscillators

Umesh Kumar Verma, Amit Sharma, Neeraj Kumar Kamal, J¨ urge n Kurths, and Manish Dev Shrimali. Explosive death induced by mean–ﬁeld diﬀusion in ide ntical oscillators. Scientiﬁc reports , 7(1):7936, 2017

2017
[34]

First order transition to oscillation death through an environme nt

Umesh Kumar Verma, Amit Sharma, Neeraj Kumar Kamal, and Man ish Dev Shri- mali. First order transition to oscillation death through an environme nt. Physics Letters A, 382(32):2122–2126, 2018

2018
[35]

Amplitude death in coupled chao tic solid-state lasers with cavity-conﬁguration-dependent instabilities

Ming-Dar Wei and Jau-Ching Lun. Amplitude death in coupled chao tic solid-state lasers with cavity-conﬁguration-dependent instabilities. Applied Physics Letters , 91(6):061121, 2007

2007
[36]

Explosive synch ronization in a general complex network

Xiyun Zhang, Xin Hu, J Kurths, and Zonghua Liu. Explosive synch ronization in a general complex network. Physical Review E , 88(1):010802, 2013

2013
[37]

Explosive death of conjugate coupled van der pol oscillators on networks

Nannan Zhao, Zhongkui Sun, Xiaoli Yang, and Wei Xu. Explosive death of conjugate coupled van der pol oscillators on networks. Physical Review E , 97(6):062203, 2018. 10

2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [7]

Critical neuromorphic computing bas ed on explo- sive synchronization

Jaesung Choi and Pilwon Kim. Critical neuromorphic computing bas ed on explo- sive synchronization. Chaos: An Interdisciplinary Journal of Nonlinear Science , 29(4):043110, 2019

work page 2019

[2] [9]

Informa- tion processing capacity of dynamical systems

Joni Dambre, David Verstraeten, Benjamin Schrauwen, and Se rge Massar. Informa- tion processing capacity of dynamical systems. Scientiﬁc reports , 2:514, 2012

work page 2012

[3] [1]

Chaotic oscillations and bifurcations in squid gian t axons

Kazuyuki Aihara. Chaotic oscillations and bifurcations in squid gian t axons. Chaos, pages 257–269, 1986

work page 1986

[4] [2]

Computation with cha os: A paradigm for cortical activity

Agnesa Babloyantz and Carlos Louren¸ co. Computation with cha os: A paradigm for cortical activity. Proceedings of the National Academy of Sciences , 91(19):9027–9031, 1994

work page 1994

[5] [3]

Neuronal avalanches in neocort ical circuits

John M Beggs and Dietmar Plenz. Neuronal avalanches in neocort ical circuits. Jour- nal of neuroscience , 23(35):11167–11177, 2003

work page 2003

[6] [4]

Being critical of criticality in the br ain

John M Beggs and Nicholas Timme. Being critical of criticality in the br ain. Frontiers in physiology , 3:163, 2012

work page 2012

[7] [5]

Marker s of criticality in phase synchronization

Maria Botcharova, Simon F Farmer, and Luc Berthouze. Marker s of criticality in phase synchronization. Frontiers in systems neuroscience , 8:176, 2014

work page 2014

[8] [6]

Nonlinear prediction of chaotic time series

Martin Casdagli. Nonlinear prediction of chaotic time series. Physica D: Nonlinear Phenomena, 35(3):335–356, 1989. 8

work page 1989