Pith. sign in

REVIEW 3 major objections 5 minor 49 references

Network structural change point detection and reconstruction for balanced neuronal networks

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that connectivity changes in balanced spiking networks can be detected from voltage fluctuations, and that segmenting recordings at those change points sharply improves wiring reconstruction.

desk verdict Genuinely useful CPD-TDCC empirical result, but the load-bearing subspace theory is wrong for the simulated matrices and the missing negative controls leave detection selectivity unproven. read the letter →

arxiv 2507.02450 v1 pith:OYX6WNLI submitted 2025-07-03 q-bio.NC

Kai Chen , Mingzhang Wang , Songting Li , Douglas Zhou This is my paper · ORCID
classification q-bio.NC
keywords networkreconstructionchangepointdetectionbalancedneuronaltime-delayedcorrelationcoefficientvoltagefluctuationstructuralplasticity
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

Structural connectivity in cortex is not static: synaptic rewiring and strength changes alter the network while recordings are being taken, and standard reconstruction methods that assume a fixed wiring pattern degrade. The paper proposes detecting the moments of structural change from voltage fluctuations alone, then reconstructing the connectivity of each stable segment with pairwise time-delayed correlation (TDCC) applied to spike trains. It argues that in balanced excitatory-inhibitory networks the recurrent input dominates the second-order voltage difference, so the fluctuation statistics carry information about the current connectivity matrix. In simulations of leaky integrate-and-fire (LIF), Hodgkin-Huxley, and Morris-Lecar networks, segmenting at detected change points raises reconstruction AUC from roughly 0.81, 0.74, and 0.68 to roughly 0.99, 0.98, and 0.94, respectively. The practical significance is that if this holds, connectivity inference from sparse experimental recordings should be performed on segments separated by detected structural changes rather than on pooled data.

What carries the argument

The load-bearing object is the second-order voltage difference $\Delta^2 v(t) = v(t+2\Delta t) - 2v(t+\Delta t) + v(t)$, decomposed into ionic, external Poisson, and recurrent contributions. Under balanced-network scalings (coupling strengths $O(1/\sqrt{K})$ and input rates $O(K)$), the recurrent term has variance of order $\sqrt{K\Delta t/N}$, which dominates the external term of order $\sqrt{\Delta t/N}$, so the population-averaged $\Delta^2 v$ inherits the direction of the recurrent input and hence of the connectivity matrix $W$. The detector is the projection magnitude $|\langle \hat{v}_n, \Delta^2 v\rangle|$ onto the smallest-singular-value right singular vector estimated from pre-change data; an F-statistic on segment variances converts the projection shift into a change-point test. The reconstruction step uses the time-delayed correlation coefficient (TDCC) between binarized spike trains, which the paper shows is proportional to the underlying coupling strength in the weak-coupling limit.

What would settle it

Run the same F-statistic detector on a balanced leaky integrate-and-fire network whose connectivity matrix does not change, and compare the peak F values with the changing-network case; if the unchanged network produces peaks above the same threshold at a comparable rate, the detected variance shift does not single out structural changes.

Watch

Extended reading notes

Core claim

The central claim is that a connectivity change $W_1 \to W_2$ at time $t_c$ produces a detectable change in the distribution of the second-order voltage difference $\Delta^2 v(t) = v(t+2\Delta t) - 2v(t+\Delta t) + v(t)$. In the balanced regime the recurrent-synaptic term of $\Delta^2 v$ dominates the external and ionic terms, so $\Delta^2 v$ lies essentially in the column space of the active connectivity matrix: $\Delta^2 v_{\mathrm{rec}}(t) \in \operatorname{col}(W_1)$ before $t_c$ and $\in \operatorname{col}(W_2)$ after. The paper argues that a vector orthogonal to $\operatorname{col}(W_1)$, estimated from pre-change data as the right singular vector of the $\Delta^2 V$ matrix with the smallest singular value, will have near-zero projection of $\Delta^2 v$ before the change and a significantly nonzero projection after, and an F-test on the variance of this projection locates the change point. Recursively segmenting the recording at such points and applying TDCC to spike trains within each segment reconstructs the successive connectivity matrices with AUC values around 0.94-0.99 in the paper's simulations, compared with 0.68-0.86 for unsegmented TDCC. The same pipeline is reported to work for changes in coupling weights with fixed topology when the weight heterogeneity exceeds a small threshold.

Load-bearing premise

The detection argument assumes that changing the wiring rotates the set of directions the voltage fluctuations can explore, so that a direction untouched before the change shows a clear response after it; if the old and new wiring produce overlapping direction sets, the stated test has nothing to detect.

Editorial extensions

If this is right

  • Wiring inference from spike trains should be run on segments separated by detected structural change points rather than on the full pooled recording.
  • The detector is claimed to work for strength changes with fixed topology, with detectability increasing monotonically with coupling-weight heterogeneity and falling off below a heterogeneity of about 0.2 in the paper's LIF tests.
  • The pipeline operates on a 200-neuron subnetwork sampled from a 4000-neuron balanced network, so it is aimed at the sparse-recording regime typical of experiments.
  • Because the change-point statistic is an F-test, the method returns a p-value for each candidate change and can be applied recursively to find multiple successive changes.

Reading between the lines

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

  • Beyond the paper: the stated subspace-separation story is not the only possible mechanism for the reported detections; for the sparse random connectivity matrices used, $W$ is generically full rank, so the empirical signal may actually come from a change in the covariance of the voltage fluctuations rather than from disjoint column spaces.
  • Beyond the paper: a direct test would be a constant-connectivity negative control; if the F-statistic peaks above threshold at a comparable rate with unchanged wiring, the variance shift is not specific to structural changes.
  • Beyond the paper: the projection-of-fluctuations idea could be extended to proxy voltage signals such as calcium imaging, which would address the paper's stated limitation that membrane-potential recordings are required.
  • Beyond the paper: since TDCC is linearly related to coupling strength in the weak-coupling limit, segmented reconstruction could in principle be used to estimate the magnitudes of synaptic weights, not just the adjacency pattern; the paper demonstrates correlation improvement but stops short of weight calibration.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript proposes a two-stage framework, CPD-TDCC, for reconstructing neuronal connectivity in balanced spiking networks that undergo abrupt structural changes. The first stage computes the second-order voltage difference Δ²v(t), projects it onto a direction \v\n estimated as the smallest right singular vector of pre-change data, and uses an F-test on consecutive windows to detect variance changes attributed to connectivity changes (Sec. 3.3, Eqs. 24–27). The second stage segments spike trains at detected change points and computes pairwise time-delayed correlation coefficients to reconstruct the adjacency matrix or coupling strengths (Secs. 2.2, 3.4–3.6). The manuscript reports AUC improvements from 0.809–0.862 to 0.985–0.990 for LIF networks, from 0.743 to 0.981 for Hodgkin-Huxley networks, and from 0.684 to 0.944 for Morris-Lecar networks, plus detection and reconstruction for heterogeneous coupling-weight changes.

Significance. If the method works as claimed, it addresses a real gap: most connectivity inference assumes stationarity, while synaptic plasticity can change structure on recording timescales. The paper has genuine strengths: the variance-dominance calculation (Eq. 22) is a parameter-free asymptotic prediction verified in Fig. 1; the TDCC-to-coupling-strength link (Eqs. 8–13) is a useful derivation; and the numerical demonstrations span three neuron models and two types of structural change. However, the central theoretical mechanism used to explain why the detector is sensitive to connectivity changes (Eq. 26) is not valid for the full-rank matrices simulated, and the paper lacks negative controls that would establish selectivity of the detector for connectivity changes as opposed to generic nonstationarity or in-sample selection effects. The empirical results are therefore promising but not yet supported by the stated mechanism.

major comments (3)
  1. [§3.3, Eq. (26), Fig. 2a–b] The subspace-separation argument is not valid for the simulated connectivity matrices. For an N×N sparse random W with in-degree K=320 and N=4000–32000, W is generically of full rank, so col(W1)=col(W2)=R^N and the two column spaces overlap completely; for the recorded 200-neuron subnetwork, the 200×N matrix is generically of rank 200, so both column spaces coincide with R^200. Hence Δ²v_rec cannot be detected by an 'orthogonal to col(W1)' vector in the sense of Eq. (26). The detector actually uses \v\n, the smallest right singular vector of the pre-change Δ²V^T, which is a finite-sample minimum-variance direction; its sensitivity must be explained through the covariance of Δ²v, specifically the variance of \v\n^T W Δd, rather than through column-space non-overlap. This distinction matters because Eq. (26) is the paper's stated justification for interpreting F-statistic peaks as connectivity-induced. The authors should either derive the covariance-based statistic or state and verify a rank-deficient regime for W.
  2. [§3.3, Eq. (27), Fig. 2e] The paper does not establish that the reported F-statistic peaks are selective for connectivity changes. Because \v\n is estimated from the first segment of the same recording that is later tested, the in-sample variance along \v\n is minimized by construction; under constant W, a later window will tend to have larger variance even without any structural change. Figure 2e shows a spurious peak at 500 ms, which the text attributes to this estimation-window effect and says can be 'eliminated by adjusting the data length' without quantifying the adjustment or the false-alarm rate. In addition, the projection time series is temporally correlated, because consecutive Δ²v samples share voltage observations, so the claimed F-distribution with d1=d2=ΔT/Δt−1 should be calibrated under the null. The manuscript should report repeated-seed simulations with constant W and with the change scenarios, giving the distribution of F-statistic peaks and detection rates at a fixed threshold, and should include a baseline comparison against a fixed random projection direction or another standard detector. Without these controls, the central claim that detected change points are connectivity-induced is not fully supported.
  3. [§3.4–3.6, Figs. 3–5] The headline AUC numbers come from a single simulation per condition. For a stochastic method with tunable windows and thresholds, reporting one realization makes it impossible to judge whether the improvements over native TDCC are stable. I request repeated-seed means and standard deviations, or bootstrapped intervals, for the AUC values in Figs. 3l–o and 4e,h and for the correlation improvements in Fig. 5k. This is not a request for a new method, but for minimal statistical reporting of the existing experiments.
minor comments (5)
  1. [Eqs. (19), (21)] In the recurrent term, [dj(t+Δt) − di(t)] should be [dj(t+Δt) − dj(t)]; the same typo appears in the third term of Eq. (21).
  2. [§2.2, Table 1] There are several typographical errors: 'causaly' should be 'causally', 'expend' should be 'expand' near Eq. (10), and Table 1's caption contains a formatting artifact ('T able 1').
  3. [§3.4, Fig. 3b] The text describes 'a sharp transition of mean values' in the projection magnitude, but Eq. (27) tests a change in variance; because the plotted quantity is |⟨\v\n, Δ²v⟩|, a variance shift changes the mean of the absolute value, but the connection should be stated explicitly to avoid confusion.
  4. [§4, Discussion] The Discussion appropriately lists limitations, but the claimed failure for external-input strength Fi=O(√K) is not tested; a single numerical experiment in that regime would make the stated scope concrete.
  5. [Declarations] The data and code availability statement ('will be made available based on reasonable requests') is weak for a simulation paper; a repository link would improve reproducibility.

Circularity Check

1 steps flagged · score 3.0 of 10

Partial in-sample circularity in the CPD statistic's projection fit; the variance-scaling derivation and TDCC reconstruction remain parameter-free and independent.

  1. fitted input called prediction [Sec. 3.3, Eq. 27, Fig. 2e]
    "In our CPD algorithm, we proxy α with the right singular vector ˆvn of Δ2V⊤ associated with the smallest singular value... We note that an additional peak appears at 500 ms, which stems from using the first 500 ms voltage times series to estimate the projection vector. This false alarm can be eliminated by adjusting the data length used for analysis."

    The test direction v_n is the right singular vector of the pre-change Δ²V^T with the smallest singular value, i.e., the in-sample minimum-variance direction over the first 500 ms. The F-statistic in Eq. 27 compares the variance of ⟨v_n, Δ²v⟩ in each window to the variance in the immediately preceding window, including that same estimation window. Because v_n minimizes the estimation-window variance by construction, the F ratio at the boundary is biased upward even if W is unchanged. The paper's own Fig. 2e displays exactly such a spurious 500 ms peak and dismisses it as an artifact of the estimation choice, showing that part of the reported detection signal is induced by fitting the projection to the test record rather than by a connectivity change.

full rationale

The main derivation (Eqs. 17–22) is a parameter-free asymptotic analysis of balanced-network voltage fluctuations, verified against simulation rather than fitted to the change-detection outcome. The TDCC-to-coupling-strength relation (Eqs. 9–13) is derived from Taylor expansion and binary-process identities, not imported from the authors' previous work; the citation to [17] is illustrative rather than load-bearing. The only circular element found is the in-sample choice of the projection direction v_n: because v_n is the minimum-variance direction of the estimation window, the F-test at the boundary is biased by construction, and the paper itself acknowledges the resulting false peak at 500 ms. This is a partial, localized circularity in the detector's statistical test, not a reduction of the central reconstruction claims, since the AUC gains are evaluated on independent post-change TDCC estimates. Separately, the subspace non-overlap assertion in Eq. 26 is mathematically questionable for full-rank sparse W matrices, but that is a correctness concern rather than a circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The framework inherits the balanced-network scaling and stationarity assumptions from prior theory and adds one paper-specific assumption, the col(W) subspace separation of Eq. 26, that is not rigorously established. The free parameters are procedural: F-test window size, projection-vector estimation length, rejection threshold, and TDCC bin and lag, each chosen per simulation without a stated rule. No new physical entities are introduced.

free parameters (4)
  • F-test window size ΔT = 20 ms (Fig. 2), 400 ms (Fig. 3), 200 ms (Fig. 4), 500 ms (Fig. 5)
    No selection criterion is given; the window size changes the F statistic and which peaks are significant, so the detection result depends on this per-simulation choice.
  • Initial projection-vector estimation window = 500 ms (Fig. 2), 10 s (Figs. 3-5)
    The length of pre-change data used to estimate v_n is set by hand per simulation; the boundary of this window produces the acknowledged false F-test peak at 500 ms in Fig. 2e.
  • F-test rejection threshold = Unspecified; detections quoted with p < 1e-20
    The critical F value that separates detection from false alarm is never stated, so thresholds are effectively set post hoc to flag the known change times in each simulation.
  • TDCC bin size and time lag (Δt, τ) = (0.1 ms, 0.1 ms) Fig. 3; (0.5 ms, 0.1 ms) Fig. 4; (0.5 ms, 0.5 ms) Fig. 5
    TDCC estimation parameters are chosen per network model without sensitivity analysis or an automated selection rule.
assumptions (5)
  • domain assumption Balanced-regime scaling: F_i ~ O(1/√K), ν̄ ~ O(K), W_ij ~ O(1/√K), m_j ~ O(1)
    Used throughout Sec. 3.2 (Eqs. 17-22) to conclude that recurrent-input fluctuations dominate external and ionic terms; restricts the theory to balanced networks with this scaling.
  • domain assumption Spike indicators within Δt behave as independent Bernoulli variables with p = m_j Δt
    Sec. 3.2, Eqs. 15-18. Ignores spike correlations across neurons and serial correlations; the central limit theorem is applied to these weakly dependent variables to obtain Gaussian fluctuation statistics.
  • ad hoc to paper Δ²v_rec lies in col(W), and col(W1) and col(W2) are non-overlapping with high probability
    Sec. 3.3, Eq. 26 and Fig. 2a. This is the load-bearing assumption for change point detection; it is not justified for full-rank random connectivity matrices, where col(W1) = col(W2) = R^N.
  • domain assumption Wide-sense stationarity of spike trains within each segment between change points
    Required by the TDCC definition (Sec. 2.2, Eq. 7) and by the F-test null hypothesis; stationarity is asserted for each segmented piece but not tested.
  • standard math Ionic currents are Lipschitz continuous
    Sec. 3.2, Eq. 20: bounds the ionic-current contribution to the second-order voltage difference as O((Δt)^2); satisfied by the LIF, Hodgkin-Huxley and Morris-Lecar models used.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Network structural change point detection and reconstruction for balanced neuronal networks." pith.science (2026). https://pith.science/paper/OYX6WNLI

@misc{pith2026250702450,
  author       = {Pith},
  title        = {Pith review of: Network structural change point detection and reconstruction for balanced neuronal networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OYX6WNLI}},
  note         = {Machine review of arXiv:2507.02450}
}
read the original abstract

Understanding brain dynamics and functions critically depends on knowledge of the network connectivity among neurons. However, the complexity of brain structural connectivity, coupled with continuous modifications driven by synaptic plasticity, makes its direct experimental measurement particularly challenging. Conventional connectivity inference methods based on neuronal recordings often assumes a static underlying structural connectivity and requires stable statistical features of neural activities, making them unsuitable for reconstructing structural connectivity that undergoes changes. To fulfill the needs of reconstructing networks undergoing potential structural changes, we propose a unified network reconstruction framework that combines connectivity-induced change point detection (CPD) with pairwise time-delayed correlation coefficient (TDCC) method. For general neuronal networks in balanced regimes, we develop a theoretical analysis for discriminating changes in structural connectivity based on the fluctuation of neuronal voltage time series. We then demonstrate a pairwise TDCC method to reconstruct the network using spike train recordings segmented at the detected change points. We show the effectiveness of our CPD-TDCC network reconstruction using large-scale network simulations with multiple neuronal models. Crucially, our method accommodates networks with changes in both network topologies and synaptic coupling strengths while retaining accuracy even with sparsely sampled subnetwork data, achieving a critical advancement for practical applications in real experimental situations. Our CPD-TDCC framework addresses the critical gap in network reconstruction by accounting connectivity-induced changes points, potentially offering a valuable tool for studying structure and dynamics in the cortical brain.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 29 canonical work pages

  1. [1]

    Nature Reviews Neuroscience 10(3), 186–198 (2009) https://doi.org/10.1038/ nrn2575

    Bullmore, E., Sporns, O.: Complex brain networks: Graph theoretical analysis of structural and functional systems. Nature Reviews Neuroscience 10(3), 186–198 (2009) https://doi.org/10.1038/ nrn2575

    work page 2009

  2. [2]

    Trends in Cognitive Sciences24(4), 302–315 (2020) https://doi.org/10.1016/j.tics.2020.01.008

    Su´ arez, L.E., Markello, R.D., Betzel, R.F., Misic, B.: Linking structure and function in macroscale brain networks. Trends in Cognitive Sciences24(4), 302–315 (2020) https://doi.org/10.1016/j.tics.2020.01.008

    work page doi:10.1016/j.tics.2020.01.008 2020

  3. [3]

    Neuroscience (2021)

    The MICrONS Consortium, Bae, J.A., Baptiste, M., Bishop, C.A., Bodor, A.L., Brittain, D., Buchanan, J., Bumbarger, D.J., Castro, M.A., Celii, B., Cobos, E., Collman, F., Da Costa, N.M., Dorkenwald, S., Elabbady, L., Fahey, P.G., Fliss, T., Froudarakis, E., Gager, J., Gamlin, C., Gray-Roncal, W., Halageri, A., Hebditch, J., Jia, Z., Joyce, E., Joyce, J., J...

    work page 2021

  4. [4]

    Nature Biotechnology 39(12), 1521–1528 (2021) https://doi.org/10.1038/s41587-021-00986-5

    Xu, F., Shen, Y., Ding, L., Yang, C.-Y., Tan, H., Wang, H., Zhu, Q., Xu, R., Wu, F., Xiao, Y., Xu, C., Li, Q., Su, P., Zhang, L.I., Dong, H.-W., Desimone, R., Xu, F., Hu, X., Lau, P.-M., Bi, G.-Q.: High- throughput mapping of a whole rhesus monkey brain at micrometer resolution. Nature Biotechnology 39(12), 1521–1528 (2021) https://doi.org/10.1038/s41587-...

    work page doi:10.1038/s41587-021-00986-5 2021

  5. [5]

    The Journal of Neuroscience 18(24), 10464– 10472 (1998) https://doi.org/10.1523/JNEUROSCI.18-24-10464.1998

    Bi, G.-q., Poo, M.-m.: Synaptic Modifications in Cultured Hippocampal Neurons: Dependence on Spike Timing, Synaptic Strength, and Postsynaptic Cell Type. The Journal of Neuroscience 18(24), 10464– 10472 (1998) https://doi.org/10.1523/JNEUROSCI.18-24-10464.1998

    arXiv 1998

  6. [6]

    Annual Review of Neuroscience 31(1), 25–46 (2008) https://doi.org/10.1146/annurev.neuro.31.060407.125639

    Caporale, N., Dan, Y.: Spike Timing–Dependent Plasticity: A Hebbian Learning Rule. Annual Review of Neuroscience 31(1), 25–46 (2008) https://doi.org/10.1146/annurev.neuro.31.060407.125639

    work page arXiv 2008

  7. [7]

    Nature Reviews Neuroscience 13(7), 478–490 (2012) https://doi.org/10.1038/nrn3258

    Caroni, P., Donato, F., Muller, D.: Structural plasticity upon learning: Regulation and functions. Nature Reviews Neuroscience 13(7), 478–490 (2012) https://doi.org/10.1038/nrn3258

    work page doi:10.1038/nrn3258 2012

  8. [8]

    Neural Computation 9(6), 1265–1275 (1997) https://doi.org/10.1162/neco

    Bedenbaugh, P., Gerstein, G.L.: Multiunit normalized cross correlation differs from the average single- unit normalized correlation. Neural Computation 9(6), 1265–1275 (1997) https://doi.org/10.1162/neco. 1997.9.6.1265

    doi:10.1162/neco 1997

Show all 49 references
  1. [9]

    Bioinformatics (Oxford, England) 18(SUPPL

    Steuer, R., Kurths, J., Daub, C.O., Weise, J., Selbig, J.: The mutual information: Detecting and evalu- ating dependencies between variables. Bioinformatics (Oxford, England) 18(SUPPL. 2), 231–240 (2002) https://doi.org/10.1093/bioinformatics/18.suppl 2.S231

    2002 doi

  2. [10]

    Physical Review E 97(5), 052216 (2018) https://doi.org/10.1103/ PhysRevE.97.052216

    Li, S., Xiao, Y., Zhou, D., Cai, D.: Causal inference in nonlinear systems: Granger causality versus time-delayed mutual information. Physical Review E 97(5), 052216 (2018) https://doi.org/10.1103/ PhysRevE.97.052216

    2018

  3. [11]

    Econometrica : journal of the Econometric Society 37(3), 424 (1969) https://doi.org/10.2307/1912791

    Granger, C.W.J.: Investigating causal relations by econometric models and cross-spectral methods. Econometrica : journal of the Econometric Society 37(3), 424 (1969) https://doi.org/10.2307/1912791

    1969 doi

  4. [12]

    In: Schelter, B., Winterhalder, M., Timmer, J

    Ding, M., Chen, Y., Bressler, S.L.: Granger causality: Basic theory and application to neuroscience. In: Schelter, B., Winterhalder, M., Timmer, J. (eds.) Handbook of Time Series Analysis, 1st edn., pp. 437–460. Wiley, ??? (2006). https://doi.org/10.1002/9783527609970.ch17

    2006 doi

  5. [13]

    Physical Review Letters 111(5), 54102 (2013) https://doi.org/10.1103/PhysRevLett

    Zhou, D., Xiao, Y., Zhang, Y., Xu, Z., Cai, D.: Causal and structural connectivity of pulse-coupled non- linear networks. Physical Review Letters 111(5), 54102 (2013) https://doi.org/10.1103/PhysRevLett. 111.054102

    2013 doi

  6. [14]

    Physical Review E86(6), 061121 (2012) https://doi.org/10.1103/PhysRevE.86.061121

    Runge, J., Heitzig, J., Marwan, N., Kurths, J.: Quantifying causal coupling strength: A lag-specific measure for multivariate time series related to transfer entropy. Physical Review E86(6), 061121 (2012) https://doi.org/10.1103/PhysRevE.86.061121

    2012 doi

  7. [15]

    Physical Review Letters 85(2), 461–464 (2000) https: //doi.org/10.1103/PhysRevLett.85.461

    Schreiber, T.: Measuring information transfer. Physical Review Letters 85(2), 461–464 (2000) https: //doi.org/10.1103/PhysRevLett.85.461

    2000 doi

  8. [16]

    Physical Review Letters 103(23), 238701 (2009) https://doi.org/10.1103/PhysRevLett.103

    Barnett, L., Barrett, A.B., Seth, A.K.: Granger causality and transfer entropy Are equivalent for gaussian variables. Physical Review Letters 103(23), 238701 (2009) https://doi.org/10.1103/PhysRevLett.103. 238701 19

    2009 doi

  9. [17]

    Proceedings of the National Academy of Sciences 121(14), 2305297121 (2024) https://doi.org/10.1073/pnas.2305297121

    Tian, Z.-q.K., Chen, K., Li, S., McLaughlin, D.W., Zhou, D.: Causal connectivity measures for pulse- output network reconstruction: Analysis and applications. Proceedings of the National Academy of Sciences 121(14), 2305297121 (2024) https://doi.org/10.1073/pnas.2305297121

    2024 doi

  10. [18]

    Biometrika 41(1/2), 100 (1954) https://doi.org/10.2307/ 2333009

    Page, E.S.: Continuous Inspection Schemes. Biometrika 41(1/2), 100 (1954) https://doi.org/10.2307/ 2333009

    1954

  11. [19]

    Biological Reviews 100(2), 892–919 (2025) https://doi.org/10.1111/brv.13167

    Hemraj, D.A., Carstensen, J.: Towards ecosystem-based techniques for tipping point detection. Biological Reviews 100(2), 892–919 (2025) https://doi.org/10.1111/brv.13167

    2025 doi

  12. [20]

    arXiv (2015)

    Pepelyshev, A., Polunchenko, A.S.: Real-Time Financial Surveillance via Quickest Change-Point Detection Methods. arXiv (2015). https://doi.org/10.48550/ARXIV.1509.01570

    2015 doi arXiv

  13. [21]

    Computer Science Review 29, 1–13 (2018) https://doi.org/10.1016/j.cosrev.2018.05.001

    Kendrick, L., Musial, K., Gabrys, B.: Change point detection in social networks—Critical review with experiments. Computer Science Review 29, 1–13 (2018) https://doi.org/10.1016/j.cosrev.2018.05.001

    2018 doi

  14. [22]

    Neurocomputing 52–54, 849–855 (2003) https://doi.org/10.1016/S0925-2312(02)00815-9

    Ratnam, R., Goense, J.B.M., Nelson, M.E.: Change-point detection in neuronal spike train activity. Neurocomputing 52–54, 849–855 (2003) https://doi.org/10.1016/S0925-2312(02)00815-9

    2003 doi

  15. [23]

    Frontiers in Systems Neuroscience 10 (2016) https://doi

    Koepcke, L., Ashida, G., Kretzberg, J.: Single and Multiple Change Point Detection in Spike Trains: Comparison of Different CUSUM Methods. Frontiers in Systems Neuroscience 10 (2016) https://doi. org/10.3389/fnsys.2016.00051

    2016

  16. [24]

    In: 2023 11th International IEEE/EMBS Conference on Neural Engineering (NER), pp

    Urzay, C., Ahad, N., Azabou, M., Schneider, A., Atamkuri, G., Hengen, K.B., Dyer, E.L.: Detecting change points in neural population activity with contrastive metric learning. In: 2023 11th International IEEE/EMBS Conference on Neural Engineering (NER), pp. 1–4. IEEE, Baltimor...

    2023

  17. [25]

    Communications Biology 8(1), 599 (2025) https://doi.org/10.1038/s42003-025-07991-3

    Marin-Llobet, A., Manasanch, A., Dalla Porta, L., Torao-Angosto, M., Sanchez-Vives, M.V.: Neural models for detection and classification of brain states and transitions. Communications Biology 8(1), 599 (2025) https://doi.org/10.1038/s42003-025-07991-3

    2025 doi

  18. [26]

    Imaging Neuroscience 3, 00520 (2025) https: //doi.org/10.1162/imag a 00520

    Ondrus, M., Olds, E., Cribben, I.: Factorized binary search: Change point detection in the network structure of multivariate high-dimensional time series. Imaging Neuroscience 3, 00520 (2025) https: //doi.org/10.1162/imag a 00520

    2025 doi

  19. [27]

    Knowledge and Information Systems 51(2), 339–367 (2017) https://doi.org/10.1007/s10115-016-0987-z

    Aminikhanghahi, S., Cook, D.J.: A survey of methods for time series change point detection. Knowledge and Information Systems 51(2), 339–367 (2017) https://doi.org/10.1007/s10115-016-0987-z

    2017 doi

  20. [28]

    In: 2016 Annual Conference on Information Science and Systems (CISS), pp

    Mosqueiro, T., Strube-Bloss, M., Tuma, R., Pinto, R., Smith, B.H., Huerta, R.: Non-parametric change point detection for spike trains. In: 2016 Annual Conference on Information Science and Systems (CISS), pp. 545–550. IEEE, Princeton, NJ (2016). https://doi.org/10.1109/CISS.20...

    2016

  21. [29]

    arXiv (2021)

    Zhang, Y., Chen, H.: Graph-Based Multiple Change-Point Detection. arXiv (2021). https://doi.org/10. 48550/ARXIV.2110.01170

    2021 doi arXiv

  22. [30]

    ACM Transactions on the Web 15(4), 1–22 (2021) https://doi.org/10.1145/3457906

    Wang, H., Qiao, C., Guo, X., Fang, L., Sha, Y., Gong, Z.: Identifying and Evaluating Anomalous Structural Change-based Nodes in Generalized Dynamic Social Networks. ACM Transactions on the Web 15(4), 1–22 (2021) https://doi.org/10.1145/3457906

    2021 doi

  23. [31]

    Fron- tiers of Engineering Management 12(1), 154–176 (2025) https://doi.org/10.1007/s42524-025-4109-z

    Xu, R., Song, Z., Wu, J., Wang, C., Zhou, S.: Change-point detection with deep learning: A review. Fron- tiers of Engineering Management 12(1), 154–176 (2025) https://doi.org/10.1007/s42524-025-4109-z

    2025 doi

  24. [32]

    Science 274(5293), 1724–1726 (1996) https://doi.org/10.1126/science.274.5293

    Van Vreeswijk, C., Sompolinsky, H.: Chaos in Neuronal Networks with Balanced Excitatory and Inhibitory Activity. Science 274(5293), 1724–1726 (1996) https://doi.org/10.1126/science.274.5293. 1724 20

    1996 doi

  25. [33]

    Neural Computation 10(6), 1321–1371 (1998) https://doi.org/10.1162/089976698300017214

    Vreeswijk, C.V., Sompolinsky, H.: Chaotic Balanced State in a Model of Cortical Circuits. Neural Computation 10(6), 1321–1371 (1998) https://doi.org/10.1162/089976698300017214

    1998 doi

  26. [34]

    Science 327(5965), 587–590 (2010) https://doi.org/10.1126/science

    Renart, A., De La Rocha, J., Bartho, P., Hollender, L., Parga, N., Reyes, A., Harris, K.D.: The Asyn- chronous State in Cortical Circuits. Science 327(5965), 587–590 (2010) https://doi.org/10.1126/science. 1179850

    2010 doi

  27. [35]

    The Journal of Neuroscience 18(10), 3870–3896 (1998) https://doi.org/10.1523/JNEUROSCI.18-10-03870.1998

    Shadlen, M.N., Newsome, W.T.: The Variable Discharge of Cortical Neurons: Implications for Connec- tivity, Computation, and Information Coding. The Journal of Neuroscience 18(10), 3870–3896 (1998) https://doi.org/10.1523/JNEUROSCI.18-10-03870.1998

    1998

  28. [36]

    The Journal of Neuroscience 26(17), 4535–4545 (2006) https://doi.org/10.1523/JNEUROSCI.5297-05.2006

    Haider, B., Duque, A., Hasenstaub, A.R., McCormick, D.A.: Neocortical Network Activity In Vivo Is Generated through a Dynamic Balance of Excitation and Inhibition. The Journal of Neuroscience 26(17), 4535–4545 (2006) https://doi.org/10.1523/JNEUROSCI.5297-05.2006

    2006 doi

  29. [37]

    Frontiers in Computational Neuroscience 14, 79 (2020) https://doi.org/10.3389/fncom.2020

    Tian, G., Li, S., Huang, T., Wu, S.: Excitation-Inhibition Balanced Neural Networks for Fast Signal Detection. Frontiers in Computational Neuroscience 14, 79 (2020) https://doi.org/10.3389/fncom.2020. 00079

    2020 doi

  30. [38]

    Physical Review Letters 134(6), 068403 (2025) https://doi.org/10.1103/ PhysRevLett.134.068403

    Barzon, G., Busiello, D.M., Nicoletti, G.: Excitation-Inhibition Balance Controls Information Encod- ing in Neural Populations. Physical Review Letters 134(6), 068403 (2025) https://doi.org/10.1103/ PhysRevLett.134.068403

    2025

  31. [39]

    Frontiers in Computational Neuroscience 12, 47 (2018)

    Gu, Q., Tian, Z., Kovacic, G., Cai, D.: The dynamics of balanced spiking neuronal networks under poisson drive is not chaotic. Frontiers in Computational Neuroscience 12, 47 (2018)

    2018

  32. [40]

    eLife 12, 86365 (2023) https: //doi.org/10.7554/eLife.86365

    Wang, C., Zhang, T., Chen, X., He, S., Li, S., Wu, S.: BrainPy, a flexible, integrative, efficient, and extensible framework for general-purpose brain dynamics programming. eLife 12, 86365 (2023) https: //doi.org/10.7554/eLife.86365

    2023 doi

  33. [41]

    The Journal of Physiology 117(4), 500–544 (1952) https://doi.org/ 10.1113/jphysiol.1952.sp004764

    Hodgkin, A.L., Huxley, A.F.: A quantitative description of membrane current and its application to conduction and excitation in nerve. The Journal of Physiology 117(4), 500–544 (1952) https://doi.org/ 10.1113/jphysiol.1952.sp004764

    1952 doi

  34. [42]

    Journal of Computational Neuroscience 27(3), 369–390 (2009) https: //doi.org/10.1007/s10827-009-0151-9

    Sun, Y., Zhou, D., Rangan, A.V., Cai, D.: Library-based numerical reduction of the Hodgkin–Huxley neuron for network simulation. Journal of Computational Neuroscience 27(3), 369–390 (2009) https: //doi.org/10.1007/s10827-009-0151-9

    2009 doi

  35. [43]

    Computational Neuroscience

    Dayan, P., Abbott, L.F.: Theoretical Neuroscience: Computational and Mathematical Modeling of Neural Systems, First paperback ed edn. Computational Neuroscience. MIT Press, Cambridge, Mass. (2005)

    2005

  36. [44]

    Biophysical Journal 35(1), 193–213 (1981) https://doi.org/10.1016/S0006-3495(81)84782-0

    Morris, C., Lecar, H.: Voltage oscillations in the barnacle giant muscle fiber. Biophysical Journal 35(1), 193–213 (1981) https://doi.org/10.1016/S0006-3495(81)84782-0

    1981 doi

  37. [45]

    PLoS ONE 6(11), 27431 (2011) https://doi.org/10.1371/journal.pone.0027431

    Ito, S., Hansen, M.E., Heiland, R., Lumsdaine, A., Litke, A.M., Beggs, J.M.: Extending Transfer Entropy Improves Identification of Effective Connectivity in a Spiking Cortical Network Model. PLoS ONE 6(11), 27431 (2011) https://doi.org/10.1371/journal.pone.0027431

    2011 doi

  38. [46]

    Scientific Reports15(1), 5357 (2025) https://doi.org/10.1038/s41598-025-88596-y

    Laasch, N., Braun, W., Knoff, L., Bielecki, J., Hilgetag, C.C.: Comparison of derivative-based and correlation-based methods to estimate effective connectivity in neural networks. Scientific Reports15(1), 5357 (2025) https://doi.org/10.1038/s41598-025-88596-y

    2025 doi

  39. [47]

    John Wiley & Sons, Inc, Hoboken, New Jersey (2014)

    Florescu, I.: Probability and Stochastic Processes. John Wiley & Sons, Inc, Hoboken, New Jersey (2014)

    2014

  40. [48]

    In: Uncertainty Proceedings 1994, pp

    Pearl, J.: A Probabilistic Calculus of Actions. In: Uncertainty Proceedings 1994, pp. 454–462. Elsevier, 21 ??? (1994). https://doi.org/10.1016/B978-1-55860-332-5.50062-6

    1994 doi

  41. [49]

    National Science Review 9(4), 228 (2022) https://doi.org/10.1093/nsr/nwab228 22

    Hou, J.-W., Ma, H.-F., He, D., Sun, J., Nie, Q., Lin, W.: Harvesting random embedding for high- frequency change-point detection in temporal complex systems. National Science Review 9(4), 228 (2022) https://doi.org/10.1093/nsr/nwab228 22

    2022 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.