Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Transient Dynamics in Lattices of Differentiating Ring Oscillators

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Lattices of differentiating neuron rings synchronize into local phase domains like Kuramoto oscillator lattices, with the steady-state domain scale set by how rings share neurons — a tunable, low-power reservoir-computing substrate.

desk verdict A real contribution on ring oscillator dynamics, but the correlation-length analysis is quantitatively misspecified and the reservoir claims are untested. read the letter →

arxiv 2506.07253 v1 pith:EEX3ZGAW submitted 2025-06-08 cs.NE

Peter DelMastro , Arjun Karuvally , Hananel Hazan , Hava Siegelmann , Edward Rietman This is my paper
classification cs.NE
keywords differentiatingneuronsringoscillatorsKuramotomodeloscillatorsynchronizationphasereductioncorrelationlengthreservoircomputingevent-basedsimulation
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

Large lattices of differentiating neurons — units whose binary output responds to the rate of change of their input rather than its level — behave, in simulation, like lattices of Kuramoto oscillators: they organize into growing regions of phase-synchronized oscillation separated by thin domain boundaries. The paper shows that the steady-state size of these regions, measured by a correlation length $\xi$, is controlled by how adjacent rings share neurons, and that the dominant oscillation mode is set by the same connectivity template. If this carries over to hardware, the lattices are a tunable, low-power reservoir-computing substrate: the neurons are dormant most of the time, so the circuits draw little steady current, yet the synchronization scale provides a handle on the network's computational regime. This is the first large-scale characterization, at tens of thousands of neurons, of a neuron class previously limited to small hand-designed networks.

What carries the argument

The central object is the ring oscillator: $n$ differentiating neurons in a cycle, each with a binary output set by an inverting Schmitt trigger acting on the derivative of its capacitor voltage, so pulses of activity circulate around the ring and no two adjacent neurons fire at once. Three mechanisms carry the argument: (1) an event-based simulator that steps the network from one output change to the next using closed-form voltage updates and treats cascades of input-driven output changes as instantaneous; (2) a numerical phase-reduction scheme that assigns each ring the pulse count $k$ of the orbit it converges to and a phase $\theta\in[0,1)$ on that orbit, with ring similarity $\langle k,\theta;k',\theta'\rangle=\cos^2((\theta-\theta')/2)$ for $k=k'$ and $0$ otherwise; and (3) the correlation length $\xi$ extracted by fitting $C(d)\propto e^{-d/\xi}$ to the mean similarity between rings at Manhattan distance $d$. Homogeneous lattices are built by reflecting a connectivity template $(L,T,R,B)$ across lattice edges so that the directed graph is $k$-colorable, which guarantees that a global periodic orbit exists; the template then determines how pulses are created, deleted, and merged at shared neurons, which is what sets the domain scale.

What would settle it

Replace the instantaneous cascade rule with a small, fixed transmission delay (and admit odd-length rings) and re-measure the steady-state correlation length for the 6-ring $(1,2,1,2)$, 8-ring $(1,3,1,3)$, and 16-ring $(4,4,4,4)$ templates: if the plateaus, the global synchronization of 4-ring lattices, or the coexistence of the $k=6$ and $k=7$ domains disappear, the domain dynamics are simulator artifacts. A breadboard or CMOS realization of a few hundred neurons wired in the same templates would settle it directly, since the prediction is that measured domain sizes follow the simulated ordering of $\xi$ across templates.

Watch

Extended reading notes

Core claim

The paper's central claim is that lattices of differentiating-neuron ring oscillators reproduce the local synchronization dynamics of the Kuramoto model. Starting from random initial states, the rings converge to a dominant periodic orbit whose pulse count $k_{\text{dom}}$ depends on the lattice connectivity template $(L,T,R,B)$ and stays near the most saturated cycle $\lfloor N/2\rfloor$; rings in that cycle form phase-coherent domains, and the thin boundaries between domains are made of rings in non-dominant cycle types. A numerical phase reduction assigns each ring a pulse count and phase $(k,\theta)$, and the correlation length $\xi$ from the fit $C(d)\propto e^{-d/\xi}$ grows over the first tens of $\tau$ and then plateaus at a template-dependent value: some templates (e.g. $(1,1,1,3)$, $(1,3,1,3)$) reach $\xi>10$ and approach global synchronization, while others saturate at short range, and 4-ring lattices synchronize globally even at 250×250 scale. The paper also establishes that an $n$-ring has exactly $\lfloor n/2\rfloor$ stable periodic orbits, one per pulse count $k$, with period given by the polynomial equation $p_{nk}(x)=v_{\text{thl}}x^n-x^{2k}+x^k-v_{\text{thl}}=0$ with $x=e^{-P/n}$, and that pulse separation grows as rings converge to their orbits.

Load-bearing premise

The lattice results rest on the event-based simulator's assumption that cascades of neuron output changes are instantaneous, with no transmission delay, and on the exclusion of odd-length rings because that instantaneity makes the cascade loop forever (Section III A); if real propagation delays or odd ring sizes alter how pulses are created, deleted, and merged, the reported domain sizes and correlation lengths could be simulation artifacts rather than properties of the circuits.

Editorial extensions

If this is right

  • Four-ring lattices reach the globally synchronized 2-cycle in finite time even at 250×250 sites, so a fully phase-locked reservoir is achievable with the smallest rings.
  • For 6-ring and 8-ring lattices the correlation length grows during the first 10–40$\tau$ and then plateaus at a value fixed by $(L,T,R,B)$, giving a finite, tunable domain scale instead of global synchronization.
  • The dominant cycle type is selected by the connectivity template and stays close to $\lfloor N/2\rfloor$, so the network's operating regime is set by wiring geometry rather than ring size alone.
  • Sixteen-ring lattices with $(4,4,4,4)$ sustain coexisting $k=6$ and $k=7$ domains for at least $10^5\tau$, a persistent spatial heterogeneity that is itself a resource for a reservoir.
  • Because the synchronizing dynamics mirror Kuramoto lattices while the circuits are mostly dormant between pulses, the lattices are a candidate low-power physical reservoir whose input–output properties could be tuned through connectivity.

Reading between the lines

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

  • A direct testable extension the paper leaves implicit: stream a time-varying input through one edge of a lattice whose template places $\xi$ near the synchronization transition, and check whether readout performance peaks there, in line with the edge-of-synchronization results reported for other oscillator reservoirs.
  • Odd-length rings are the natural first probe of whether the reported dynamics are physical: the simulator excludes them only because instantaneous cascades can loop forever, so a simulator or circuit with finite transmission delay would either confirm that pulse creation, deletion, and merging are unchanged or reveal that the domain structure is partly an artifact.
  • The similarity metric zeroes out pairs of rings with different pulse counts; a graded metric would show whether domain boundaries are sharp or diffuse and could connect the observed boundaries to the vortices and defects of two-dimensional Kuramoto lattices.
  • The paper's concluding proposal of mixed integrator-differentiator networks suggests an experiment: add a small fraction of integrator neurons to a synchronized lattice and measure whether the correlation-length plateau survives, i.e. whether neuronal heterogeneity shifts the tunable range of $\xi$.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. This manuscript studies networks of differentiating neurons organized into rings and coupled into rectangular lattices. The authors introduce an event-based simulation, characterize the stable periodic orbits of isolated rings (claiming one stable orbit per pulse count and deriving a polynomial for the oscillation period), and report that homogeneous lattices develop locally synchronized phase domains whose steady-state size, quantified by a correlation length xi, depends on the ring-sharing template (L,T,R,B). They propose that these tunable lattices could serve as low-power reservoir computers.

Significance. The analytical period equation (Eq. 23) and the stable-cycle characterization are useful contributions to a sparsely studied class of neural circuits, and the qualitative observation of domain formation in coupled differentiating ring oscillators is intriguing. The event-based algorithm is described in enough detail to reimplement, which is a strength. If the correlation-length analysis can be put on a sound footing, the demonstration that synchronization scale is controlled by coupling geometry would be a valuable, falsifiable design principle for neuromorphic substrates. At present, however, the quantitative claim rests on a misspecified correlation metric and an unvalidated simulator approximation.

major comments (3)
  1. [V C, Eqs. (6)-(7), (24)-(25)] The phase theta is defined as a normalized variable in [0,1) in Eq. (6), so the statement in Eq. (24) that complete anti-phase corresponds to |theta-theta'|=pi is inconsistent with that definition; with the stated normalization a half-cycle phase difference is 0.5, and cos^2((theta-theta')/2) would give approximately 0.94, not 0. If theta is instead meant to be in radians, the phase-reduction definition must be changed accordingly. Independently, for independent uniformly distributed phases the expectation of cos^2((theta-theta')/2) is 1/2, so C(d) in Eq. (25) does not decay to zero; fitting C(d) proportional to exp(-d/xi) without subtracting this disconnected baseline flattens the log slope and can produce spuriously large or template-dependent xi values. The xi ordering in Fig. 10 is therefore not a reliable measure of true correlation decay. Please redefine the metric consistently and fit the connected correlation, e.g., <cos(theta-theta')> or cos^2(pi(theta-theta')) with the 1/2 baseline removed.
  2. [V C and Fig. 10] The correlation-length fit is reported without the distance range used, the number of fitted points, standard errors on xi, or any goodness-of-fit measure. Given the baseline issue, the differences between templates may reflect the fitting range or noise rather than physics. Please provide fit diagnostics and error bars on xi, and state exactly which d values are included in the regression.
  3. [III A] All lattice results come from the event-based simulator in Algorithm 1, in which cascades propagate instantaneously and odd-length cycles are excluded because Step 3 of Algorithm 1 can sometimes enter an infinite loop. The paper does not validate this approximation against direct numerical integration of Eq. (4) or against hardware measurements with finite transmission delays. Since pulse creation, deletion, and merging (Fig. 5) is exactly what controls domain growth, the reported correlation lengths and domain sizes are conditional on the zero-delay approximation; please add a sensitivity check (e.g., a small-lattice ODE comparison or a finite-delay variant) and discuss the odd-cycle caveat in the conclusions.
minor comments (4)
  1. [II C] The sentence 'We will show that we show below that for differentiating ring oscillator networks are similar to Kuramoto lattices' contains a duplicated phrase and a grammar error; please rewrite.
  2. [V A] The text says the ring at site (i,j) shares L_ij nodes with 'the ring to its right (site(i-1,j))' and then defines R_ij with site (i,j+1); the indices for left and right neighbors are inconsistent and should be corrected.
  3. [V D and Fig. 10 caption] 'One standard derivation' should be 'one standard deviation', and 'correlation rate saturates' should presumably be 'correlation length saturates'.
  4. [IV A] The claim that only floor(n/2) orbits are stable attractors is presented as a simulation finding; please specify the stability criterion and the initial-condition sampling used, since the phase reduction in Section V C relies on this claim.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: the ring-orbit and lattice-correlation results come from the paper's own simulator and are not fitted to the targeted conclusions; the only notable self-citation (ref. 22) is background and not load-bearing.

full rationale

The central claim that homogeneous lattices of differentiating ring oscillators develop local phase correlations whose steady-state scale depends on the sharing template is supported by direct simulation measurements (Figs. 8-10), not by any parameter fitted to that claim. The period equation (Eqs. 19-23) is a self-consistency condition derived from the observed k/n duty-cycle structure and the Schmitt-trigger threshold v_thl; it is not used to define the number of stable orbits, which is an independent simulation finding, and it is not fitted to any target output. The phase-reduction procedure of Section V C is operational and applied to simulation states. The correlation length xi in Eq. 25 and Fig. 10 is a descriptive fit to simulation output, not an input to the model, so no 'prediction' reduces to a fitted parameter by construction. The paper cites the authors' prior hardware work (ref. 22) for the existence of ring oscillations and for the necklace count, but these are background; the simulations independently reproduce the relevant ring dynamics and even correct ref. 22's stable-orbit count. No load-bearing step reduces to a self-citation chain or to a definitional equivalence. A separate concern is that the cos^2 similarity metric in Eq. 24 has a 1/2 disconnected baseline for independent random phases, so the exponential fit for xi may be misspecified and the reported xi values could be inflated. That is a statistical correctness risk, not circularity, because xi is a measurement of simulation output rather than a parameter that determines that output.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

Free parameters are the unreported thresholds and the fitted xi values. Additional axioms are modeling choices and empirical assertions that the period and phase analysis depend on; none of these assumptions encode the main result, but several are not fully specified or proven.

free parameters (2)
  • Schmitt trigger thresholds v_thl and v_thh = not reported in the paper
    Thresholds enter the period equation (Eq. 22) and every simulation; without their values, the reported periods and domain dynamics cannot be reproduced exactly.
  • Correlation length xi for each (L,T,R,B) template = e.g., xi > 10 for (1,1,1,3) and (1,3,1,3); smaller for other templates; no confidence intervals
    Estimated by least-squares fit of log C(d) to -d/xi (Section V C, Fig. 10); this fitted quantity is the quantitative evidence for the claim that domain scale depends on connectivity.
assumptions (6)
  • domain assumption Output cascades after an event are computed instantaneously, with no transmission delay.
    Section III A models cascading effects as occurring instantaneously. All lattice simulations assume this; finite delay is acknowledged to change odd-ring behavior.
  • domain assumption Only output states with no adjacent 1s are allowed, enforced by the Schmitt trigger and binary outputs.
    Section II A and Section III A. This restriction defines the state space and affects pulse propagation; it comes from hardware but is an assumed model.
  • domain assumption In the stable k-cycle, pulses are uniformly spaced and each neuron has duty cycle k/n, so the three-phase period derivation applies.
    Section IV B before Eq. 22. The authors state this was observed, but it is assumed in deriving the period polynomial.
  • domain assumption The polynomial p_nk has at most two roots in [0,1], and the larger root is unstable, so k/n uniquely determines the stable period.
    Section IV B: 'We've found empirically...' The uniqueness is needed for the phase reduction used in lattices but is not proven.
  • domain assumption Each ring state encountered during coupled-lattice simulation converges to a k-cycle, so the asymptotic phase map gamma is well-defined.
    Section V C assigns phase by simulating convergence to the periodic orbit; strong coupling can in principle push rings outside the basin, and no check is reported.
  • standard math The homogeneous reflection construction guarantees a global periodic orbit for the lattice.
    Section V B uses a graph coloring argument; this is standard but ensures the homogeneous lattices studied are a special, non-generic class.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transient Dynamics in Lattices of Differentiating Ring Oscillators." pith.science (2026). https://pith.science/paper/EEX3ZGAW

@misc{pith2026250607253,
  author       = {Pith},
  title        = {Pith review of: Transient Dynamics in Lattices of Differentiating Ring Oscillators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EEX3ZGAW}},
  note         = {Machine review of arXiv:2506.07253}
}
read the original abstract

Recurrent neural networks (RNNs) are machine learning models widely used for learning temporal relationships. Current state-of-the-art RNNs use integrating or spiking neurons -- two classes of computing units whose outputs depend directly on their internal states -- and accordingly there is a wealth of literature characterizing the behavior of large networks built from these neurons. On the other hand, past research on differentiating neurons, whose outputs are computed from the derivatives of their internal states, remains limited to small hand-designed networks with fewer than one-hundred neurons. Here we show via numerical simulation that large lattices of differentiating neuron rings exhibit local neural synchronization behavior found in the Kuramoto model of interacting oscillators. We begin by characterizing the periodic orbits of uncoupled rings, herein called ring oscillators. We then show the emergence of local correlations between oscillators that grow over time when these rings are coupled together into lattices. As the correlation length grows, transient dynamics arise in which large regions of the lattice settle to the same periodic orbit, and thin domain boundaries separate adjacent, out-of-phase regions. The steady-state scale of these correlated regions depends on how the neurons are shared between adjacent rings, which suggests that lattices of differentiating ring oscillator might be tuned to be used as reservoir computers. Coupled with their simple circuit design and potential for low-power consumption, differentiating neural nets therefore represent a promising substrate for neuromorphic computing that will enable low-power AI applications.

Figures

Figures reproduced from arXiv: 2506.07253 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_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8 [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9 [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10 [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]

Discussion (0). Sign in 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

  1. Reservoir Computation with Networks of Differentiating Neuron Ring Oscillators

    cs.NE 2025-07 conditional novelty 6.0 of 10

    Small-world networks of differentiating neuron ring oscillators achieve 90.65% accuracy on MNIST and are proposed as an energy-efficient reservoir computing substrate.

Reference graph

Works this paper leans on

46 extracted references · 34 canonical work pages · cited by 1 Pith paper

  1. [1]

    author author J. L. \ Elman ,\ title title Finding Structure in Time , \ https://doi.org/10.1207/s15516709cog1402\_1 journal journal Cognitive Science \ volume 14 ,\ pages 179--211 ( year 1990 ) NoStop

    doi:10.1207/s15516709cog1402 1990

  2. [2]

    Hochreiter \ and\ author J

    author author S. Hochreiter \ and\ author J. Schmidhuber ,\ title title Long short-term memory , \ @noop journal journal Neural Computation \ ( year 1997 ) NoStop

    work page 1997

  3. [3]

    author author N. B. \ Erichson , author O. Azencot , author A. F. \ Queiruga , author L. Hodgkinson , \ and\ author M. W. \ Mahoney ,\ title title Lipschitz recurrent neural networks , \ in\ https://openreview.net/forum?id=-N7PBXqOUJZ booktitle 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021 \ ( p...

    work page 2021

  4. [4]

    author author T. K. \ Rusch \ and\ author S. Mishra ,\ title title Coupled oscillatory recurrent neural network (cornn): An accurate and (gradient) stable architecture for learning long time dependencies , \ in\ https://openreview.net/forum?id=F3s69XzWOia booktitle 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria...

    work page 2021

  5. [5]

    author author T. K. \ Rusch , author S. Mishra , author N. B. \ Erichson , \ and\ author M. W. \ Mahoney ,\ title title Long expressive memory for sequence modeling , \ in\ https://openreview.net/forum?id=vwj6aUeocyf booktitle The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022 \ ( publisher OpenRevi...

    work page 2022

  6. [6]

    Gu , author K

    author author A. Gu , author K. Goel , \ and\ author C. R \' e ,\ title title Efficiently modeling long sequences with structured state spaces , \ in\ https://openreview.net/forum?id=uYLFoz1vlAC booktitle The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022 \ ( publisher OpenReview.net ,\ year 2022 ) NoStop

    work page 2022

  7. [7]

    Chandra , author S

    author author R. Chandra , author S. Goyal , \ and\ author R. Gupta ,\ title title Evaluation of deep learning models for multi-step ahead time series prediction , \ @noop journal journal IEEE Access \ volume 9 ,\ pages 83105--83123 ( year 2021 ) NoStop

    work page 2021

  8. [8]

    Pathak , author B

    author author J. Pathak , author B. R. \ Hunt , author M. Girvan , author Z. Lu , \ and\ author E. Ott ,\ title title Model-free prediction of large spatiotemporally chaotic systems from data: A reservoir computing approach. \ @noop journal journal Physical review letters \ volume 120 2 ,\ pages 024102 ( year 2018 ) NoStop

    work page 2018

Show all 46 references
  1. [9]

    Melis , author C

    author author G. Melis , author C. Dyer , \ and\ author P. Blunsom ,\ title title On the state of the art of evaluation in neural language models , \ @noop journal journal ArXiv \ volume abs/1707.05589 ( year 2017 ) NoStop

    2017 arXiv

  2. [10]

    Merity , author N

    author author S. Merity , author N. S. \ Keskar , \ and\ author R. Socher ,\ title title Regularizing and optimizing lstm language models , \ @noop journal journal ArXiv \ volume abs/1708.02182 ( year 2017 ) NoStop

    2017 arXiv

  3. [11]

    Gu \ and\ author T

    author author A. Gu \ and\ author T. Dao ,\ title title Mamba: Linear-time sequence modeling with selective state spaces , \ @noop journal journal ArXiv \ volume abs/2312.00752 ( year 2023 ) NoStop

    2023 arXiv

  4. [12]

    Meng , author R

    author author L. Meng , author R. B. \ Gorbet , \ and\ author D. Kuli'c ,\ title title Memory-based deep reinforcement learning for pomdps , \ @noop journal journal 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) \ ,\ pages 5619--5626 ( year 202...

    2021

  5. [13]

    Jaeger , author M

    author author H. Jaeger , author M. Lukoševičius , author D. Popovici , \ and\ author U. Siewert ,\ title title 2007 special issue: Optimization and applications of echo state networks with leaky- integrator neurons , \ https://api.semanticscholar.org/CorpusID:17795421 journal...

    2007

  6. [14]

    author author J. J. \ Hopfield ,\ title title Neural networks and physical systems with emergent collective computational abilities. \ @noop journal journal Proceedings of the National Academy of Sciences of the United States of America \ volume 79 8 ,\ pages 2554--8 ( year 19...

    1982

  7. [15]

    Krotov \ and\ author J

    author author D. Krotov \ and\ author J. J. \ Hopfield ,\ title title Large associative memory problem in neurobiology and machine learning , \ in\ https://openreview.net/forum?id=X4y\_10OX-hX booktitle 9th International Conference on Learning Representations, ICLR 2021, Virtu...

    2021

  8. [16]

    Karuvally , author T

    author author A. Karuvally , author T. J. \ Sejnowski , \ and\ author H. T. \ Siegelmann ,\ title title Energy-based general sequential episodic memory networks at the adiabatic limit , \ 10.48550/ARXIV.2212.05563 journal journal CoRR \ volume abs/2212.05563 ( year 2022 ),\ 10...

    doi arXiv

  9. [17]

    Doya ,\ title title Bifurcations of recurrent neural networks in gradient descent learning , \ @noop journal journal IEEE Transactions on Neural Networks \ ( year 1993 ) NoStop

    author author K. Doya ,\ title title Bifurcations of recurrent neural networks in gradient descent learning , \ @noop journal journal IEEE Transactions on Neural Networks \ ( year 1993 ) NoStop

    1993

  10. [18]

    Bengio , author P

    author author Y. Bengio , author P. Simard , \ and\ author P. Frasconi ,\ title title Learning long-term dependencies with gradient descent is difficult , \ 10.1109/72.279181 journal journal IEEE transactions on neural networks / a publication of the IEEE Neural Networks Counc...

    doi

  11. [19]

    Pascanu , author T

    author author R. Pascanu , author T. Mikolov , \ and\ author Y. Bengio ,\ title title On the difficulty of training recurrent neural networks , \ in\ https://api.semanticscholar.org/CorpusID:14650762 booktitle International Conference on Machine Learning \ ( year 2012 ) NoStop

    2012

  12. [20]

    author author K. M. \ Stiefel \ and\ author J. S. \ Coggan ,\ title title The energy challenges of artificial superintelligence , \ https://api.semanticscholar.org/CorpusID:264497615 journal journal Frontiers in Artificial Intelligence \ volume 6 ( year 2023 ) NoStop

    2023

  13. [21]

    Hasslacher \ and\ author M

    author author B. Hasslacher \ and\ author M. W. \ Tilden ,\ title title Theoretical foundations for nervous networks , \ 10.1063/1.54209 journal journal AIP Conference Proceedings \ volume 411 ,\ pages 179--184 ( year 1997 ) NoStop

    1997 doi

  14. [22]

    author author E. A. \ Rietman , author M. W. \ Tilden , \ and\ author M. Askenazi ,\ title title Analog computation with rings of quasiperiodic oscillators: the microdynamics of cognition in living machines , \ @noop journal journal Robotics Auton. Syst. \ volume 45 ,\ pages 2...

    2003

  15. [23]

    Frigo \ and\ author M

    author author J. Frigo \ and\ author M. W. \ Tilden ,\ title title Satbot i: Prototype of a biomorphic autonomous spacecraft , \ https://www.osti.gov/biblio/150940 \ ( year 1995 ) NoStop

    1995

  16. [24]

    author author J. R. \ Frigo \ and\ author M. W. \ Tilden ,\ title title Biologically inspired neural network controller for an infrared tracking system , \ in\ 10.1117/12.335726 booktitle Mobile Robots XIII and Intelligent Transportation Systems, Boston, MA, USA, November 1, 1...

    doi

  17. [25]

    Hasslacher \ and\ author M

    author author B. Hasslacher \ and\ author M. W. \ Tilden ,\ title title Living machines , \ 10.1016/0921-8890(95)00019-C journal journal Robotics Auton. Syst. \ volume 15 ,\ pages 143--169 ( year 1995 ) NoStop

    1995 doi

  18. [26]

    Still \ and\ author M

    author author S. Still \ and\ author M. W. \ Tilden ,\ title Controller for a four-legged walking machine , \ in\ 10.1142/9789812816535\_0012 booktitle Neuromorphic Systems \ ( year 1998 )\ pp.\ pages 138--148 ,\ http://arxiv.org/abs/https://www.worldscientific.com/doi/pdf/10....

    1998 doi

  19. [27]

    Rodan \ and\ author P

    author author A. Rodan \ and\ author P. Tiňo ,\ title title Minimum complexity echo state network , \ https://api.semanticscholar.org/CorpusID:319654 journal journal IEEE Transactions on Neural Networks \ volume 22 ,\ pages 131--144 ( year 2011 ) NoStop

    2011

  20. [28]

    author author J. P. \ Sethna ,\ title Statistical mechanics: Entropy, order parameters, and complexity , \ \ ( publisher Oxford Press ,\ year 2021 )\ Chap.\ chapter 10: Correlations, response, and dissipation ,\ edition 2nd \ ed. Stop

    2021

  21. [29]

    author author H. Nakao ,\ title title Phase reduction approach to synchronisation of nonlinear oscillators , \ @noop journal journal Contemporary Physics \ volume 57 ,\ pages 188 -- 214 ( year 2016 ) NoStop

    2016

  22. [30]

    author author Y. Kuramoto ,\ title title Self-entrainment of a population of coupled non-linear oscillators , \ in\ @noop booktitle International Symposium on Mathematical Problems in Theoretical Physics ,\ editor edited by\ editor H. Araki \ ( publisher Springer Berlin Heidel...

    1975

  23. [31]

    Kuramoto ,\ title title Chemical oscillations, waves, and turbulence , \ in\ @noop booktitle Springer Series in Synergetics \ ( year 1984 ) NoStop

    author author Y. Kuramoto ,\ title title Chemical oscillations, waves, and turbulence , \ in\ @noop booktitle Springer Series in Synergetics \ ( year 1984 ) NoStop

    1984

  24. [32]

    author author J. A. \ Acebr \'o n , author L. L. \ Bonilla , author C. J. P. \ Vicente , author F. Ritort , \ and\ author R. Spigler ,\ title title The kuramoto model: A simple paradigm for synchronization phenomena , \ @noop journal journal Reviews of Modern Physics \ volume ...

    2005

  25. [33]

    author author F. A. \ Rodrigues , author T. K. D. \ Peron , author P. Ji , \ and\ author J. Kurths ,\ title title The kuramoto model in complex networks , \ https://doi.org/10.1016/j.physrep.2015.10.008 journal journal Physics Reports \ volume 610 ,\ pages 1--98 ( year 2016 ) ...

    2015 doi

  26. [34]

    Sarkar \ and\ author N

    author author M. Sarkar \ and\ author N. Gupte ,\ title title Phase synchronization in the two-dimensional kuramoto model: Vortices and duality. \ @noop journal journal Physical review. E \ volume 103 3-1 ,\ pages 032204 ( year 2021 ) NoStop

    2021

  27. [35]

    Sarkar ,\ title title Synchronization transition in the two-dimensional kuramoto model with dichotomous noise

    author author M. Sarkar ,\ title title Synchronization transition in the two-dimensional kuramoto model with dichotomous noise. \ https://api.semanticscholar.org/CorpusID:236155269 journal journal Chaos \ volume 31 8 ,\ pages 083102 ( year 2021 ) NoStop

    2021

  28. [36]

    Maass , author T

    author author W. Maass , author T. Natschl \"a ger , \ and\ author H. Markram ,\ title title Real-time computing without stable states: A new framework for neural computation based on perturbations , \ https://api.semanticscholar.org/CorpusID:1045112 journal journal Neural Com...

    2002

  29. [37]

    Jaeger ,\ title title The''echo state''approach to analysing and training recurrent neural networks , \ \ ( year 2001 ) NoStop

    author author H. Jaeger ,\ title title The''echo state''approach to analysing and training recurrent neural networks , \ \ ( year 2001 ) NoStop

    2001

  30. [38]

    Choi \ and\ author P

    author author J. Choi \ and\ author P. Kim ,\ title title Critical neuromorphic computing based on explosive synchronization , \ http://arxiv.org/abs/1810.10944 journal journal CoRR \ volume abs/1810.10944 ( year 2018 ) ,\ http://arxiv.org/abs/1810.10944 1810.10944 NoStop

    2018 arXiv

  31. [39]

    Choi \ and\ author P

    author author J. Choi \ and\ author P. Kim ,\ title title Reservoir computing based on quenched chaos , \ http://arxiv.org/abs/1909.01571 journal journal CoRR \ volume abs/1909.01571 ( year 2019 ) ,\ http://arxiv.org/abs/1909.01571 1909.01571 NoStop

    1909 arXiv

  32. [40]

    author author R. A. \ Legenstein \ and\ author W. Maass ,\ title title Edge of chaos and prediction of computational performance for neural circuit models , \ https://api.semanticscholar.org/CorpusID:7249094 journal journal Neural networks : the official journal of the Interna...

    2007

  33. [41]

    Baranvok \ and\ author I

    author author P. Baranvok \ and\ author I. Farkas ,\ title title Memory capacity of input-driven echo state networks at the edge of chaos , \ in\ https://api.semanticscholar.org/CorpusID:12748419 booktitle International Conference on Artificial Neural Networks \ ( year 2014 ) NoStop

    2014

  34. [42]

    Dambre , author D

    author author J. Dambre , author D. Verstraeten , author B. Schrauwen , \ and\ author S. Massar ,\ title title Information processing capacity of dynamical systems , \ https://api.semanticscholar.org/CorpusID:7342429 journal journal Scientific Reports \ volume 2 ( year 2012 ) NoStop

    2012

  35. [43]

    author author C. G. \ Langton ,\ title title Computation at the edge of chaos: Phase transitions and emergent computation , \ https://api.semanticscholar.org/CorpusID:62374310 journal journal Physica D: Nonlinear Phenomena \ volume 42 ,\ pages 12--37 ( year 1990 ) NoStop

    1990

  36. [44]

    Perez-Nieves , author V

    author author N. Perez-Nieves , author V. C. H. \ Leung , author P. L. \ Dragotti , \ and\ author D. F. M. \ Goodman ,\ title title Neural heterogeneity promotes robust learning , \ @noop journal journal Nature Communications \ volume 12 ( year 2020 ) NoStop

    2020

  37. [45]

    author author T. Q. \ Chen , author Y. Rubanova , author J. Bettencourt , \ and\ author D. K. \ Duvenaud ,\ title title Neural ordinary differential equations , \ in\ @noop booktitle Neural Information Processing Systems \ ( year 2018 ) NoStop

    2018

  38. [46]

    author author K. O. \ Stanley \ and\ author R. Miikkulainen ,\ title title Evolving neural networks through augmenting topologies , \ @noop journal journal Evolutionary Computation \ volume 10 ,\ pages 99--127 ( year 2002 ) NoStop

    2002

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.