pith. sign in

arxiv: 2606.25874 · v1 · pith:AFNXGQ3Onew · submitted 2026-06-24 · 🧬 q-bio.NC · nlin.AO

Topology-Dependent Emergence of Polychronous Neuronal Groups: A Recurrence-Plot Characterization

Lucas A. T. X. Carneiro , Armand D. Jiofack , Fernando F. Ferreira This is my paper

Pith reviewed 2026-06-25 19:21 UTC · model grok-4.3

classification 🧬 q-bio.NC nlin.AO
keywords polychronous neuronal groupsWatts-Strogatz topologyclustering coefficientrecurrence plotsSTDPIzhikevich neuronssmall-world networksspike-timing-dependent plasticity
0
0 comments X

The pith

Clustering coefficient in small-world networks drives polychronous neuronal group formation, with higher clustering producing over 90 percent more groups than random graphs.

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

The paper simulates recurrent networks of 1000 Izhikevich neurons with STDP and heterogeneous delays to test how network topology affects the emergence of polychronous neuronal groups. A Watts-Strogatz sweep shows that clustering coefficient is the main driver: ring lattices with C around 0.35 yield roughly 850 PNGs while random graphs with C around 0.20 yield fewer than 50, cutting representational capacity by more than 90 percent. The authors also present a recurrence-plot method that detects these groups as diagonal structures without using neuron labels and reports high determinism in the resulting trajectories.

Core claim

A parametric Watts-Strogatz topology sweep reveals that the clustering coefficient C is the primary structural driver of PNG yield: the transition from a ring-lattice (C~0.35, ∼850 PNGs) to a random graph (C~0.20, <50 PNGs) reduces representational capacity by more than 90%. The sparse-dot-product Recurrence Plot framework identifies PNGs as unit-slope diagonal structures in the phase-space recurrence matrix, entirely independent of anatomical neuron labelling, with recurrence quantification analysis yielding DET~0.65.

What carries the argument

Watts-Strogatz parametric sweep on clustering coefficient combined with the sparse-dot-product recurrence plot decoder for label-free PNG identification.

If this is right

  • Small-world topology with elevated clustering maximizes the number of stable polychronous groups available for computation.
  • Recurrence plots supply a label-free decoder that extracts PNGs directly from the phase-space trajectory.
  • Changes in clustering coefficient can alter network representational capacity by more than an order of magnitude.
  • Determinism values near 0.65 quantify the reproducibility of the polychronous dynamics under STDP.

Where Pith is reading between the lines

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

  • If clustering is the dominant factor, biological circuits may tune local connectivity density to control the size of their polychronous repertoire.
  • The recurrence-plot approach could be tested on experimental multi-electrode recordings to detect PNGs when anatomical connectivity is unknown.
  • Repeating the sweep with larger networks or different neuron models would test whether the 90-percent reduction generalizes beyond N=1000 Izhikevich cells.

Load-bearing premise

The offline event-driven detection algorithm correctly and exhaustively identifies all true PNGs stabilized by STDP and heterogeneous delays without bias across the topology sweep.

What would settle it

Applying an independent PNG detection method to the same simulated spike trains and finding no systematic drop in counted groups when moving from high-clustering to low-clustering topologies would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.25874 by Armand D. Jiofack, Fernando F. Ferreira, Lucas A. T. X. Carneiro.

Figure 1
Figure 1. Figure 1: presents the macroscopic outcome of the ten-hour simulation. The mean firing-rate time series ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Excitatory synaptic weight distributions. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Representative detected [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Statistical characterisation of the 1 545 unique [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Topology–polychronization phase diagram ( [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Joint raster (top) and recurrence matrix [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
read the original abstract

Polychronous Neuronal Groups (PNGs) reproducible, time-locked spatiotemporal firing cascades stabilised by Spike-Timing-Dependent Plasticity (STDP) and heterogeneous axonal delays provide a combinatorially rich substrate for neural computation whose structural determinants remain poorly understood. We simulate a recurrent network of N=1000 Izhikevich neurons over ten hours of biological time and identify 1545 unique PNGs via an offline event-driven detection algorithm. A parametric Watts-Strogatz topology sweep reveals that the clusteringcoefficient C is the primary structural driver of PNG yield: the transition from a ring-lattice (C~0.35, $\sim\!850$ \PNGs) to a random graph (C~!0.20$, $<\!50$ \PNGs) reduces representational capacity by more than 90%. We further introduce a sparse-dot-product Recurrence Plot (RP) framework that identifies PNGs as unit-slope diagonal structures in the phase-space recurrence matrix, entirely independent of anatomical neuron labelling. Recurrence Quantification Analysis yields DET~0.65, quantifying the reproducibility of the network's dynamical trajectory. Together, the results establish small-world topology as the structural optimum for polychronization and the \RP decoder as a principled, label-free tool for PNG identification.

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.

Referee Report

2 major / 2 minor

Summary. The manuscript simulates a recurrent network of N=1000 Izhikevich neurons with STDP and heterogeneous delays over 10 hours of biological time, identifies 1545 unique PNGs via an offline event-driven detection algorithm, and performs a parametric Watts-Strogatz topology sweep. It claims that the clustering coefficient C is the primary structural driver of PNG yield, with a ring lattice (C≈0.35) producing ~850 PNGs versus a random graph (C≈0.20) producing <50 PNGs (a >90% reduction). The work also introduces a sparse-dot-product recurrence-plot (RP) framework that detects PNGs as unit-slope diagonals in the phase-space recurrence matrix, independent of neuron labels, and reports DET≈0.65 from recurrence quantification analysis.

Significance. If the central claims hold after addressing the noted issues, the results would link small-world topology to enhanced polychronization capacity and supply a label-free RP-based decoder for PNGs. The simulation scale and explicit topology sweep constitute a concrete, falsifiable test of structural determinants; the RP approach offers a potentially general tool for identifying reproducible spatiotemporal patterns.

major comments (2)
  1. [Abstract / topology sweep] Abstract and topology-sweep results: the claim that C is the 'primary structural driver' cannot be isolated because the Watts-Strogatz rewiring probability p monotonically decreases both C and average shortest-path length L. No partial-correlation analysis, regression controlling for L, or auxiliary topologies that hold L fixed while varying C are described. Since the offline PNG detector identifies time-locked cascades whose formation can depend on global reachability, the observed 90% drop cannot be attributed primarily to C.
  2. [Methods / detection algorithm] Methods / detection algorithm and results: the reported PNG counts (1545 total, ~850 vs <50 across topologies) lack any statement of the number of independent runs, statistical tests, error bars, sensitivity to STDP parameters, or validation that the event-driven detector recovers ground-truth groups. These omissions leave the quantitative claims without robustness assessment.
minor comments (2)
  1. [Abstract] Abstract contains typographical errors: 'clusteringcoefficient', 'C~!0.20$', and inconsistent spacing around '~' and '!' symbols.
  2. [RP framework] The RP framework is introduced as 'entirely independent of anatomical neuron labelling,' but the manuscript does not explicitly compare its PNG detections against the event-driven algorithm on the same trajectories.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments. We respond to each major point below.

read point-by-point responses

  1. Referee: [Abstract / topology sweep] Abstract and topology-sweep results: the claim that C is the 'primary structural driver' cannot be isolated because the Watts-Strogatz rewiring probability p monotonically decreases both C and average shortest-path length L. No partial-correlation analysis, regression controlling for L, or auxiliary topologies that hold L fixed while varying C are described. Since the offline PNG detector identifies time-locked cascades whose formation can depend on global reachability, the observed 90% drop cannot be attributed primarily to C.

    Authors: We agree that the Watts-Strogatz model confounds C and L, so the attribution of PNG yield primarily to C is not isolated by the current sweep. In the reported data, PNG counts track C closely while L reaches near-minimal values early in the rewiring range, but this does not constitute rigorous separation. We will revise the abstract and results to describe C as a key (rather than primary) driver, add explicit discussion of the L confound, and note that future work with L-controlled topologies would be needed to strengthen the claim. revision: partial

  2. Referee: [Methods / detection algorithm] Methods / detection algorithm and results: the reported PNG counts (1545 total, ~850 vs <50 across topologies) lack any statement of the number of independent runs, statistical tests, error bars, sensitivity to STDP parameters, or validation that the event-driven detector recovers ground-truth groups. These omissions leave the quantitative claims without robustness assessment.

    Authors: We will add the missing information: the PNG counts are means over 10 independent runs per topology (with standard deviations shown as error bars in revised figures); we will include ANOVA results across topologies. A supplementary analysis of sensitivity to STDP parameters (A+, A−, τ) will be provided. We will also add a methods subsection reporting validation of the event-driven detector on synthetic spike trains containing injected ground-truth PNGs, including recovery rates. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper obtains its central result (C as primary driver of PNG yield) from explicit forward simulation of an Izhikevich network under a Watts-Strogatz parameter sweep, followed by offline event-driven PNG detection; no parameters are fitted to observed PNG counts, no equation or definition equates the counted yield to the input topology measure by construction, and no load-bearing uniqueness theorem or ansatz is imported via self-citation. The derivation chain therefore remains independent of its own outputs.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The central claim rests on the assumption that the chosen simulation scale, duration, and detection procedure faithfully capture PNG emergence; these choices function as free parameters whose values are stated but not justified against alternatives.

free parameters (3)
  • N=1000
    Network size selected for the recurrent simulation; no justification given for this scale versus smaller or larger networks.
  • simulation duration = 10 hours biological time
    Duration chosen to allow PNG stabilization; no data shown on convergence or shorter/longer runs.
  • Watts-Strogatz rewiring probability and delay distribution parameters
    Topology and delay heterogeneity parameters swept or set but specific values and sensitivity not reported in abstract.
axioms (2)
  • standard math Izhikevich neuron model equations govern single-cell dynamics
    Standard reduced neuron model invoked without re-derivation.
  • domain assumption STDP together with heterogeneous axonal delays can stabilize reproducible PNGs
    Core premise taken from prior polychronization literature and used to interpret the simulation output.

pith-pipeline@v0.9.1-grok · 5781 in / 1657 out tokens · 50591 ms · 2026-06-25T19:21:18.599677+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 1 canonical work pages

  1. [1]

    Gerstner, W

    W. Gerstner, W. M. Kistler, R. Naud, and L. Paninski,Neuronal Dynamics: From Single Neurons to Networks and Models of Cognition(Cambridge University Press, 2014)

    2014

  2. [2]

    Breakspear, Nature Neuroscience20, 340 (2017)

    M. Breakspear, Nature Neuroscience20, 340 (2017)

    2017

  3. [3]

    Z. F. Mainen and T. J. Sejnowski, Science268, 1503 (1995)

    1995

  4. [4]

    J. J. Hopfield, Nature376, 33 (1995)

    1995

  5. [5]

    E. M. Izhikevich, Neural Computation18, 245 (2006)

    2006

  6. [6]

    Spiking manifesto,

    E. M. Izhikevich, “Spiking manifesto,” (2025), arXiv:2512.11843 [cs.NE]

    work page arXiv 2025

  7. [7]

    A. P. Vaz, J. H. Wittig, S. K. Inati, and K. A. Zaghloul, Nature Communications14, 4723 (2023)

    2023

  8. [8]

    W. Xie, J. H. Wittig, J. I. Chapeton, M. El-Kalliny, S. N. Jackson, S. K. Inati, and K. A. Zaghloul, Nature635, 935 (2024)

    2024

  9. [9]

    Martinez and H

    R. Martinez and H. Paugam-Moisy, inArtificial Neural Net- works – ICANN 2009, edited by C. Alippi, M. Polycarpou, C. Panayiotou, and G. Ellinas (Springer Berlin Heidelberg, Berlin, Heidelberg, 2009) pp. 75–84

    2009

  10. [10]

    Chrol-Cannon, Y

    J. Chrol-Cannon, Y . Jin, and A. Gr¨uning, Neurocomputing267, 644 (2017)

    2017

  11. [11]

    Bullmore and O

    E. Bullmore and O. Sporns, Nature Reviews Neuroscience10, 186 (2009)

    2009

  12. [12]

    Sporns, D

    O. Sporns, D. R. Chialvo, M. Kaiser, and C. C. Hilgetag, Trends in Cognitive Sciences8, 418 (2004)

    2004

  13. [13]

    D. J. Watts and S. H. Strogatz, Nature393, 440 (1998)

    1998

  14. [14]

    Pauli, P

    R. Pauli, P. Weidel, S. Kunkel, and A. Morrison, Frontiers in Neuroinformatics12, 46 (2018)

    2018

  15. [15]

    A. L. Hodgkin and A. F. Huxley, The Journal of Physiology117, 500 (1952)

    1952

  16. [16]

    E. M. Izhikevich, IEEE Transactions on Neural Networks14, 1569 (2003)

    2003

  17. [17]

    B. L. Sabatini and W. G. Regehr, Annual Review of Physiology 61, 521 (1999)

    1999

  18. [18]

    Bi and M.-m

    G.-q. Bi and M.-m. Poo, Journal of Neuroscience18, 10464 (1998)

    1998

  19. [19]

    Markram, J

    H. Markram, J. L¨ubke, M. Frotscher, and B. Sakmann, Science 275, 213 (1997)

    1997

  20. [20]

    G. G. Turrigiano, K. R. Leslie, N. S. Desai, L. C. Rutherford, and S. B. Nelson, Nature391, 892 (1998)

    1998

  21. [21]

    Erd˝os and A

    P. Erd˝os and A. R´enyi, Publ. Math. Inst. Hung. Acad. Sci.5, 17 (1960)

    1960

  22. [22]

    R. J. Douglas and K. A. Martin, Annual Review of Neuroscience 27, 419 (2004)

    2004

  23. [23]

    V . B. Mountcastle, Brain120, 701 (1997)

    1997

  24. [24]

    Eckmann, S

    J.-P. Eckmann, S. O. Kamphorst, and D. Ruelle, Europhysics Letters4, 973 (1987)

    1987

  25. [25]

    Marwan, M

    N. Marwan, M. C. Romano, M. Thiel, and J. Kurths, Physics Reports438, 237 (2007)

    2007

  26. [26]

    Stimberg, R

    M. Stimberg, R. Brette, and D. F. M. Goodman, eLife8, e47314 (2019)

    2019

  27. [27]

    Destexhe, M

    A. Destexhe, M. Rudolph, and D. Par ´e, Nature Reviews Neuroscience4, 739 (2003)

    2003

  28. [28]

    van Vreeswijk and H

    C. van Vreeswijk and H. Sompolinsky, Science274, 1724 (1996)

    1996

  29. [29]

    Brunel, Journal of Computational Neuroscience8, 183 (2000)

    N. Brunel, Journal of Computational Neuroscience8, 183 (2000)

    2000

  30. [30]

    V ´ertes and T

    P. V ´ertes and T. Duke, HFSP Journal4, 153 (2010)

    2010

  31. [31]

    Cocchi, L

    L. Cocchi, L. L. Gollo, A. Zalesky, and M. Breakspear, Progress in Neurobiology158, 132 (2017)

    2017

  32. [32]

    G. Ma, R. Yan, and H. Tang, Patterns4, 100831 (2023)

    2023

  33. [33]

    Destexhe and M

    A. Destexhe and M. Rudolph-Lilith,Neuronal Noise, Springer Series in Computational Neuroscience (Springer US, 2012). 9 Appendix A: Brian2 network simulation code The following listing reproduces the complete Brian2 Python implementation of the baselineSNNdescribed in Sec. III A. Appendix B: OfflinePNGdetection algorithm The following listing reproduces th...

    2012