Pith.

Other Statistics · arXiv:2611.16427 v1 · submitted 12 Nov 2026 · 45 references

Exact power spectra and effective dimensionality in critical block-GOE networks

Konstantinos Grammenos, Dionissios Hristopulos

The paper says

“Criticality in neural networks is crucial for optimal information processing. Recent studies have shifted focus from static criticality to frequency-dependent behavior, particularly in oscillatory systems modeled by block-Gaussian orthogonal ensemble (block-GOE) matrices.”

more of the abstract

However, analytical tools to predict spectral properties such as power spectrum and effective dimensionality remain scarce. In this work, we derive exact solutions for the network-averaged power spectrum TrN[S(ω)], the network autocovariance TrN[C(τ)], and the normalized effective dimensionality DN(ω) in block-GOE ensembles. These results provide a comprehensive framework for understanding how oscillatory networks operate near criticality, offering insights into the dynamic balance between stability and flexibility in large-scale neural systems.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up. A genuinely analytical result for block-GOE spectral properties that had been numerical/empirical before. The derivations are careful, the scaling exponents are new, and the Fourier duality preservation is a neat observation. The paper is what it claims to be. This line is wrong

4 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 45 references say.

4 minor objections

The paper says, at Title and Abstract

Pith says: minor objection 1 of 4

The term 'exact' in the title could be misinterpreted as applying to finite N; the text clarifies the large-N limit, but the authors may wish to add an explicit qualifier (e.g., 'in the large-N limit') in the abstract for clarity.

The paper says, at Section II, Eq. (4)

Pith says: minor objection 2 of 4

The sign convention for the block +βJ and −βJ is stated but not motivated; a brief justification (e.g., referencing the Wilson–Cowan feedback loop) would help readers less familiar with the model.

show 2 more

The paper says, at Figure 2 caption

Pith says: minor objection 3 of 4

The caption reads 'Blue (dots): Euler-Maruyama simulation' – minor typos; should be 'Euler–Maruyama' and 'Broken line' should be 'dashed'.

The paper says, at Throughout

Pith says: minor objection 4 of 4

The notation η→1⁻ and η→1^− is used inconsistently; the LaTeX ^− appears in some places. This should be unified for readability.

What would settle it

Run a simulation for a very small κ (e.g., κ=0.01) at the edge of instability (η=0.999) and measure the NAPS exponent near ω_C; compare to the κ→0 limit derived in Appendix B4. If the exponent remains -1/2, the derivation holds across parameter space.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up confidence high
  2. Skeptic agrees 41 s · 31,882 tokens read, 5,190 written
  3. Editor wrote the verdict above 13 s · 32,627 tokens read, 1,382 written
  4. Referee accept: 4 minor 27 s · 32,107 tokens read, 3,306 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

This paper delivers what it promises: exact closed-form expressions for the network-averaged power spectrum, autocovariance, and normalized effective dimensionality in block-GOE networks. The novelty is real — prior work (Calvo et al.) was descriptive and numerical, and Hu & Sompolinsky covered i.i.d. Gaussian matrices only. The block-GOE case had no analytic treatment until this paper.

The derivations are mathematically serious. They use Fourier transforms, Plancherel-Parseval, Bessel functions, and confluent hypergeometric functions in a way that is clearly laid out in the appendices. The scaling exponents (-½، 3/2، -½) are derived and validated numerically with R² = 1.00 fits, and the intermediate asymptotic analysis (distinguished limit) explaining how the smooth peak turns into a singularity at the edge of instability is a nice piece of singular perturbation theory.

The assumptions are standard for this line of work: linear dynamics near equilibrium, additive white Gaussian noise, and block-GOE connectivity. The authors state these clearly and do not overclaim generality. The weakest assumption — linearization of the nonlinear dynamics — is explicitly acknowledged and is not buried.

The only soft spots are inherent to the scope. The results are for infinite network size; finite-N corrections are not examined, but that is a natural sequel rather than a flaw. The noise is uncorrelated white Gaussian; colored or correlated inputs would require a separate treatment. And the paper is heavy on special functions (Humbert Φ1 and modified Bessel functions), so it will not be light reading for an applied audience, but the critical-regime simplifications are accessible.

For whom is this? People working on random neural networks, spectral theory of large random matrices in neuroscience, or frequency-dependent criticality. It provides a analytical toolkit for oscillatory criticality that was missing.

I would accept this for a good journal as is. The analysis is thorough; the claims follow from the mathematics; the numerics support the theory. A serious referee will have checks to do but should find no load-bearing flaw.

The strongest objection

The paper derives closed-form expressions for the network-averaged power spectrum, normalized effective dimensionality, and network-averaged autocovariance within the stated assumptions (linear dynamics, additive white Gaussian noise, block-GOE connectivity). The mathematical derivations in Appendices B–D appear rigorous, using standard techniques (Plancherel-Parseval, semicircle law, confluent hypergeometric functions). The scaling laws are validated by numerical simulations for finite N. The weakest assumption—linearization near equilibrium—is acknowledged by the authors and is standard for this line of work. No internal inconsistency or unsupported leap was found. The central claim is well-supported given the scope.

The paper's central argument is internally consistent and its assumptions are explicitly stated. The derivations are mathematically detailed and the numerical evidence supports them. The reader correctly identified the linearization as the weakest assumption, which the paper acknowledges. I find no additional load-bearing concern. The concrete test suggested is a robustness check for extreme parameter values, not a fix for a flaw.

The referee's report

The work provides the first analytic treatment of these spectral and temporal quantities for the oscillatory block-GOE ensemble, moving beyond the descriptive stage and offering a predictive analytical framework. The scaling laws differ qualitatively from those of i.i.d. Gaussian networks (e.g., slower NAPS divergence, preserved Fourier duality), which has direct consequences for interpreting frequency-dependent criticality in neural data. The derivation is rigorous, uses standard random-matrix and Fourier techniques, and is fully documented in appendices.

Pith's reading of the paper (a machine's summary, not the authors')

The authors study linear neural networks driven by noise, where neurons are split into two populations (excitatory and inhibitory) with random connections described by block-Gaussian orthogonal ensemble (block-GOE) matrices. They derive exact formulas for three key observables in the infinite-neuron limit. First, the network-averaged power spectrum (NAPS) shows a singularity at a characteristic frequency when the system is at the edge of instability, scaling as the inverse square root of the distance from that frequency. Second, the normalized effective dimensionality (NED) collapses to zero at the same frequency, scaling as the 3/2 power of the distance, indicating highly coordinated activity. Third, the network-averaged autocovariance (NAAC) decays as the square root of time lag at criticality, showing critical slowing down with superimposed oscillations. These results match numerical simulations. The block-GOE network preserves the Fourier transform duality between NAPS and NAAC even at criticality, unlike previously studied i.i.d. Gaussian networks.

Core claim. In the large-N limit, the network-averaged power spectrum Tr_N[S(ω)] near the edge of instability scales as |ω - ω_C|^{-1/2} (Lemma 1), the normalized effective dimensionality D_N(ω) scales as |ω - ω_C|^{3/2} (Lemma 3), and the network-averaged autocovariance decays as τ^{-1/2} with oscillations at frequency κ (Lemma 4).

The premise it rests on. The linearized dynamics ẋ = -(I - gW)x + ξ(t) (Eq. 1) with additive white Gaussian noise and block-GOE connectivity (Eq. 4) accurately captures the essential behavior of nonlinear neural networks near equilibrium for the purpose of understanding frequency-dependent criticality. The paper acknowledges this is a linearization of Eq. (2) and that real neural dynamics include saturating nonlinearities, which may alter behavior at or beyond the edge of instability.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Criticality in neural networks is crucial for optimal information processing. Recent studies have shifted focus from static criticality to frequency-dependent behavior, particularly in oscillatory systems modeled by block-Gaussian orthogonal ensemble (block-GOE) matrices. However, analytical tools to predict spectral properties such as power spectrum and effective dimensionality remain scarce. In this work, we derive exact solutions for the network-averaged power spectrum TrN[S(ω)], the network autocovariance TrN[C(τ)], and the normalized effective dimensionality DN(ω) in block-GOE ensembles. These results provide a comprehensive framework for understanding how oscillatory networks operate near criticality, offering insights into the dynamic balance between stability and flexibility in large-scale neural systems.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 0.81

“These results provide a comprehensive framework for understanding how oscillatory networks operate near criticality, offering insights into the dynamic balance between stability and flexibility in large-scale neural systems.”

How the main claim is stated: asserted 0.90

What the abstract itself states, as probabilities: claims first 0.09 · states limitation 0.11 · human participants 0.02 · controlled comparison 0.02 · randomized assignment 0.01 · states quantitative result 0.11 · compares with prior results 0.13 · mentions code or data availability 0.02

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Criticality in neural networks is crucial for optimal information processing. background 1.00 0.03 0.03 asserted 1.00
Recent studies have shifted focus from static criticality to frequency-dependent behavior, particularly in oscillatory systems modeled by block-Gaussian orthogonal ensemble (block-GOE) matrices. background 1.00 0.04 0.03 asserted 1.00
However, analytical tools to predict spectral properties such as power spectrum and effective dimensionality remain scarce. gap 1.00 0.04 0.03 asserted 1.00
In this work, we derive exact solutions for the network-averaged power spectrum $\operatorname{Tr}_{N}[\mathbf{S}(\omega)]$, the network autocovariance $\operatorname{Tr}_{N}[\mathbf{C}(\tau)]$, and the normalized effective dimensionality $D_N(\omega)$ in block-GOE ensembles. contribution 0.92 0.98 0.38 asserted 1.00
These results provide a comprehensive framework for understanding how oscillatory networks operate near criticality, offering insights into the dynamic balance between stability and flexibility in large-scale neural systems. interpretation 1.00 0.49 0.70 asserted 0.90

What kind of paper the classifier reports this is: theoretical result 1.00 · evidence from proof or derivation 1.00

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 4,371 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (6)
Figure 1 of the paper
Figure 1. also shows least-squares fits of the NAPS val￾ues against the power-law function f(ε) = q± |ε| ν for the two branches around ωC . The linear (due to the double logarithms) fits are restricted to 100 points of each frequency interval that are closer to ωC , where the NAPS curves approximate straight lines. In the right branch (ω > ωC ), the fit produces an exponent ν+ = −0.5005 ± 3.39 × 10−5 and an amplitude q+ = 0.2149, with a coefficient of determination R2 = 1.00. Simi￾larly, on the left branch (ω < ωC ), we obtain ν− = −0.5001 ± 7.99 × 10−6 and q− = 0.918 (R2 = 1.00). Both fitted…
Figure 2 of the paper
Figure 2. FIG. 2. NAPS as a function of
Figure 3 of the paper
Figure 3. FIG. 3. Normalized effective dimensionality for
Figure 4 of the paper
Figure 4. is a visual validation of Lemma 3. To vali￾date the scaling behavior near the algebraic zero point ωC = κ, we performed a least-squares fit to the power￾law form f(ω) = d±|ω −ωC | ν separately on both sides of the transition. The same discretization of the frequency space is used as in
Figure 5 of the paper
Figure 5. FIG. 5. NAAC lim
Figure 6 of the paper
Figure 6. FIG. 6. NAAC lim

Figures are the authors', reproduced from the paper.

References (45)
  1. [1] R. Calvo, C. Martorell, G. B. Morales, S. Di Santo, and M. A. Mu˜ noz, Frequency-Dependent Covariance Reveals Critical Spatiotemporal Patterns of Synchronized Activ- ity in the Human Brain, Physical Review Letters133, 208401 (2024) on Pith
  2. [2] P. Bak,How Nature Works(Oxford University Press, London, England, 1997) on Pith
  3. [3] J. O’Byrne and K. Jerbi, How critical is brain criticality?, Trends in Neurosciences45, 820 (2022) on Pith
  4. [4] In statistical physics, equations of motion that are driven by deterministic drift and stochastic terms are known as Langevin equations on Pith
  5. [5] G. E. Uhlenbeck and L. S. Ornstein, On the Theory of the Brownian Motion, Phys. Rev.36, 823 (1930) on Pith
  6. [6] C. Gardiner,Handbook of Stochastic Methods for Physics, Chemistry, and the Natural Sciences(Springer, 1985) on Pith
  7. [7] R. Calvo, C. Martorell, A. Roig, and M. A. Mu˜ noz, Ro- bust scaling in human brain dynamics despite correlated inputs and limited sampling distortions, Physical Review Letters136, 068402 (2026), arXiv:2506.03640 [q-bio.NC] on PitharXiv
  8. [8] Y. Hu and H. Sompolinsky, The spectrum of covariance matrices of randomly connected recurrent neuronal net- works with linear dynamics, PLoS Computational Biol- ogy18, e1010327 (2022) on Pith
  9. [9] A. M. Lyapunov,General Problem of the Stability of Mo- tion, edited by A. T. Fuller (Taylor & Francis, London, 1992) on Pith
  10. [10] H. R. Wilson and J. D. Cowan, A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue, Kybernetik13, 55 (1973) on Pith
  11. [11] The connectivity type we cite in Eq. (4) is not strictly excitatory-inhibitory due to the sign ofJelements on Pith
  12. [12] For brevity we use the same symbol for the average over the steady state. The steady-state average differs from the ensemble noise average used forξ(t) on Pith
  13. [13] E. S. Meckes and M. W. Meckes, Rates of Convergence for Empirical Spectral Measures: A Soft Approach, inCon- vexity and Concentration, edited by E. Carlen, M. Madi- man, and E. M. Werner (Springer New York, New York, NY, 2017) pp. 157–181 on Pith
  14. [14] A. Guionnet and O. Zeitouni, Concentration of the Spec- tral Measure for Large Matrices, Electron. Commun. Probab.5, 119 on Pith
  15. [15] D. G. Clark, L. Abbott, and A. Litwin-Kumar, Dimen- sion of Activity in Random Neural Networks, Physical Review Letters131, 118401 (2023) on Pith
  16. [16] D. J. Higham, An Algorithmic Introduction to Numerical Simulation of Stochastic Differen- tial Equations, SIAM Review43, 525 (2001), https://doi.org/10.1137/S0036144500378302 on Pithdoi:10.1137/s0036144500378302
  17. [17] P. Welch, The use of fast Fourier transform for the esti- mation of power spectra: A method based on time aver- aging over short, modified periodograms, IEEE Transac- tions on Audio and Electroacoustics15, 70 (1967) on Pith
  18. [18] M. Abramowitz and I. A. Stegun,Handbook of Mathe- matical Functions with Formulas, Graphs, and Mathe- matical Tables, ninth printing ed. (Dover Publications, New York, 1964) on Pith
  19. [19] W. Magnus, F. Oberhettinger, and R. P. Soni,Formulas and Theorems for the Special Functions of Mathematical Physics(Springer, Berlin, Heidelberg, 1966) on Pith
  20. [20] M. Botcharova, S. F. Farmer, and L. Berthouze, Markers of criticality in phase synchronization, Frontiers in Sys- tems Neuroscience8, 10.3389/fnsys.2014.00176 (2014) on Pithdoi:10.3389/fnsys.2014.00176
show 25 more
  1. [21] J. D. Murray, A. Bernacchia, D. J. Freedman, R. Romo, J. D. Wallis, X. Cai, C. Padoa-Schioppa, T. Pasternak, H. Seo, D. Lee, and X.-J. Wang, A hierarchy of intrinsic timescales across primate cortex, Nat Neurosci17, 1661 (2014) on Pith
  2. [22] R. Gao, R. L. van den Brink, T. Pfeffer, and B. Voytek, Neuronal timescales are functionally dynamic and shaped by cortical microarchitecture, Elife9, e61277 (2020) on Pith
  3. [23] R. Hardstone, S.-S. Poil, G. Schiavone, R. Jansen, V. V. Nikulin, H. D. Mansvelder, and K. Linkenkaer-Hansen, Detrended Fluctuation Analysis: A Scale-Free View on Neuronal Oscillations, Frontiers in PhysiologyV olume 3 - 2012, 10.3389/fphys.2012.00450 (2012) on Pithdoi:10.3389/fphys.2012.00450
  4. [24] K. Linkenkaer-Hansen, V. V. Nikouline, J. M. Palva, and R. J. Ilmoniemi, Long-range temporal correlations and scaling behavior in human brain oscillations, J Neurosci 21, 1370 (2001) on Pith
  5. [25] K. J. Miller, L. B. Sorensen, J. G. Ojemann, and M. den Nijs, Power-Law Scaling in the Brain Surface Electric Potential, PLOS Computational Biology5, 1 (2009). 21 on Pith
  6. [26] B. J. He, J. M. Zempel, A. Z. Snyder, and M. E. Raichle, The temporal structures and functional significance of scale-free brain activity, Neuron66, 353 (2010) on Pith
  7. [27] F. Lombardi, O. Shriki, H. J. Herrmann, and L. de Ar- cangelis, Long-range temporal correlations in the broad- band resting state activity of the human brain revealed by neuronal avalanches, Neurocomputing461, 657 (2021) on Pith
  8. [28] L. Grafakos,Classical Fourier Analysis, Graduate Texts in Mathematics, Vol. 249 (Springer, New York, NY, 2014) on Pith
  9. [29] A. Erd´ elyi, W. Magnus, F. Oberhettinger, and F. G. Tri- comi,Tables of Integral Transforms. Vol. I, Based, in part, on notes left by Harry Bateman (McGraw-Hill Book Company, New York, Toronto, London, 1954) staff of the Bateman Manuscript Project on Pith
  10. [30] G. N. Watson,A Treatise on the Theory of Bessel Func- tions, 2nd ed. (Cambridge University Press, 1944) on Pith
  11. [31] G. A. Campbell and R. M. Foster,Fourier Integrals for Practical Applications(D. Van Nostrand Company, 1948) on Pith
  12. [32] G. I. Barenblatt,Scaling, Self-similarity, and Intermedi- ate Asymptotics: Dimensional Analysis and Intermediate Asymptotics, Cambridge Texts in Applied Mathematics, Vol. 14 (Cambridge University Press, Cambridge, 1996) on Pith
  13. [33] C. M. Bender and S. A. Orszag,Advanced Mathemati- cal Methods for Scientists and Engineers I: Asymptotic Methods and Perturbation Theory(Springer, New York, NY, 1999) on Pith
  14. [34] I. S. Gradshteyn and I. M. Ryzhik,Table of Integrals, Series, and Products, 7th ed., edited by A. Jeffrey and D. Zwillinger (Academic Press, Amsterdam, 2007) on Pith
  15. [35] A. V. Oppenheim and A. S. Willsky,Signals and Systems, 2nd ed. (Pearson, Upper Saddle River, NJ, 1996) on Pith
  16. [36] A. D. Poularikas,The Transforms and Applications Handbook: Second Edition(CRC Press, 2000) on Pith
  17. [37] M. Arshad, M. Usman, M. Z. Iqbal, and A. Ali, K- Humbert Confluent Hypergeometric Functions Φ 1,k , Φ2,k , and Φ 3,k, J. Math. Comput. Sci.35, 348 (2024) on Pith
  18. [38] A. Erdelyi, ed.,Higher Transcendental Functions, Vol. 1 (McGraw-Hill, New York, NY, 1953) on Pith
  19. [39] H. Buchholz,The Confluent Hypergeometric Function, edited by B. D. Coleman, R. Aris, L. Collatz, J. L. Erick- sen, P. Germain, M. E. Gurtin, M. M. Schiffer, E. Stern- berg, and C. Truesdell, Springer Tracts in Natural Phi- losophy, Vol. 15 (Springer, Berlin, Heidelberg, 1969) on Pith
  20. [40] A. Erd´ elyi, W. Magnus, F. Oberhettinger, and F. G. Tri- comi,Higher Transcendental Functions, Volume 2, edited by H. Bateman (McGraw-Hill, New York, 1953) based, in part, on notes left by Harry Bateman. Prepared by the Staff of the Bateman Manuscript Project on Pith
  21. [41] Z. Bai and J. W. Silverstein,Spectral Analysis of Large Dimensional Random Matrices, 2nd ed., Springer series in statistics (Springer, New York, NY, 2009) on Pith
  22. [42] H. von Storch and F. W. Zwiers,Statistical Analysis in Climate Research(Cambridge University Press, Cam- bridge, England, 2002) on Pith
  23. [43] X. Shen and Y. Hu, Covariance spectrum in nonlinear recurrent neural networks, arXiv 10.48550/arXiv.2508.05288 (2025), arXiv:2508.05288 [q-bio.NC] on PithPith's reviewarXivdoi:10.48550/arxiv.2508.05288
  24. [44] E. M. Stein and R. Shakarchi,Fourier Analysis: An In- troduction(Princeton University Press, 2011) on Pith
  25. [45] P.-C. Hang, L. Hu, and M.-J. Luo, Asymptotic expan- sions of the Humbert Function Φ 1 and their applications (2026), arXiv:2504.09280 [math.CA] on PitharXiv

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment