pith. machine review for the scientific record. sign in

arxiv: 2604.18523 · v1 · submitted 2026-04-20 · ❄️ cond-mat.dis-nn · cs.IT· math.IT· math.ST· stat.TH

Recognition: unknown

BBP transition and the leading eigenvector of the spiked Wigner model with inhomogeneous noise

Leonardo S. Ferreira , Fernando L. Metz
Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:49 UTC · model grok-4.3

classification ❄️ cond-mat.dis-nn cs.ITmath.ITmath.STstat.TH
keywords spiked Wigner modelBBP transitioninhomogeneous noiseoutlier eigenvectorhigh-dimensional inferencespectral edgespower-law distributionsignal detection
0
0 comments X

The pith

The BBP transition line becomes non-monotonic for truncated power-law noise variances, allowing inhomogeneous noise to enhance signal detectability in the spiked Wigner model.

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

This paper analyzes the eigenvalues and eigenvectors of a high-dimensional matrix that combines a rank-one signal spike with additive noise whose variances vary randomly across entries. It obtains closed equations that locate the spectral edges, any outlier eigenvalue outside the bulk, and the probability distribution of the components of the associated eigenvector. These equations fix the location of the BBP transition, which separates regimes where the spike can be detected from those where it cannot. When the variances follow a truncated power-law distribution, the transition curve bends back, indicating that moderate inhomogeneity in the noise can make detection easier than in the homogeneous case. In the detectable regime the eigenvector components furnish a direct estimator for the unknown spike.

Core claim

We derive exact equations for the spectral edges, the outlier eigenvalue, and the distribution of the components of the outlier eigenvector of an inhomogeneous rank-one spiked Wigner model. These equations determine the BBP transition line separating the gapped phase, where the signal is detectable, from the gapless phase. For a noise matrix with variances generated from a truncated power-law distribution, the BBP transition line is non-monotonic, showing that an inhomogeneous noise can enhance signal detectability. In the gapped regime, the distribution of the outlier eigenvector provides a natural estimator of the spike.

What carries the argument

Self-consistent equations for the spectral edges, outlier eigenvalue, and the distribution of outlier eigenvector components that close in the high-dimensional limit for i.i.d. variance profiles.

If this is right

  • The distribution of the outlier eigenvector components serves as an estimator for the spike vector in the gapped phase.
  • The BBP transition line can be located exactly once the variance distribution is specified.
  • For truncated power-law variances, moderate increases in inhomogeneity can move the transition boundary to permit detection at weaker signal strengths.
  • No outlier eigenvalue separates from the bulk spectrum in the gapless phase, rendering the spike undetectable.

Where Pith is reading between the lines

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

  • Accounting for entry-wise variance variation in real data sets could raise detection thresholds beyond what homogeneous-noise models predict.
  • The same non-monotonic enhancement might appear for other heavy-tailed variance distributions used in statistics or machine learning.
  • Finite-size corrections derived from the exact equations would indicate how large a matrix must be before the non-monotonicity becomes observable.

Load-bearing premise

The noise variances are drawn independently and identically from the truncated power-law distribution in the infinite-matrix-size limit with a rank-one spike.

What would settle it

Numerical diagonalization of large finite matrices with truncated power-law variances to check whether the observed onset of an outlier eigenvalue traces the predicted non-monotonic BBP curve.

Figures

Figures reproduced from arXiv: 2604.18523 by Fernando L. Metz, Leonardo S. Ferreira.

Figure 1
Figure 1. Figure 1: Spectrum of the homogeneous spiked Wigner model with [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Empirical spectral density ρ(λ) of the observation matrix An (see Eq. (1)) with a truncated power-law variance profile PS(s) defined in Eq. (32). Solid lines are the solutions of Eqs. (11) and (12) with ε = 10−3 , for different values of the width parameter ∆ and exponent α, as indicated in the figures. Dashed lines represent the Wigner semicircle law with σ = 1 + ∆ 2 (see Eq. (4)). Symbols are results obt… view at source ↗
Figure 3
Figure 3. Figure 3: Spectral edge and spectral gap of the observation matrix [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Distribution of the eigenvector components outside the bulk of eigenvalues, [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Mean squared error MSE between the spike vector Xn and the eigenvector Rn(λout) associated to the outlier eigenvalue of the observation matrix An, for both deterministic and Gaussian-distributed spikes. The variance profile PS(s) follows a truncated power-law distribution with ∆ = 1 and α = 0.5. The BBP threshold γc = γc (∆,α) is computed from the solutions of Eqs. (15) and (21). Solid lines represent anal… view at source ↗
Figure 6
Figure 6. Figure 6: Critical value γc of the signal-to-noise ratio marking the BBP transition (see the main text). When γ > γc , the spectral gap of the observation matrix is finite and the rank-one signal can be detected. These results are obtained from the solutions of Eqs. (15) and (21) for ε = 10−3 and a truncated power-law distribution PS(s) with width ∆ and shape parameter α (see Eq. (32)). The spike is generated from t… view at source ↗
Figure 7
Figure 7. Figure 7: Critical value αc of the shape parameter marking the BBP transition (see the main text). These results are obtained from the solutions of Eqs. (15) and (21) for ε = 10−3 and a truncated power-law distribution PS(s) with width ∆ and shape parameter α (see Eq. (32)). When α < αc , the spectral gap of the observation matrix is finite and the rank-one signal can be detected. The spike is generated from the dis… view at source ↗
read the original abstract

The spiked Wigner ensemble is a prototypical model for high-dimensional inference. We study the spectral properties of an inhomogeneous rank-one spiked Wigner model in which the variance of each entry of the noise matrix is itself a random variable. In the high-dimensional limit, we derive exact equations for the spectral edges, the outlier eigenvalue, and the distribution of the components of the outlier eigenvector. These equations determine the BBP transition line that separates the gapped phase, where the signal is detectable, from the gapless phase. In the gapped regime, the distribution of the outlier eigenvector provides a natural estimator of the spike. We solve the equations for a noise matrix whose variances are generated from a truncated power-law distribution. In this case, the BBP transition line is non-monotonic, showing that an inhomogeneous noise can enhance signal detectability.

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 paper analyzes the spiked Wigner model with inhomogeneous noise, where each entry's variance is drawn i.i.d. from a truncated power-law distribution. In the high-dimensional limit, it derives exact self-consistent equations governing the spectral edges, the position of the outlier eigenvalue, and the full distribution of the components of the leading eigenvector. These equations are used to locate the BBP transition separating the gapped (detectable) and gapless phases; for the truncated power-law case the transition line is shown to be non-monotonic, indicating that variance heterogeneity can improve signal detectability relative to the homogeneous case. In the gapped regime the outlier eigenvector is argued to furnish a natural estimator of the underlying spike.

Significance. If the derivations are rigorous, the work provides a concrete, solvable extension of the classical BBP transition to random variance profiles and demonstrates a counter-intuitive enhancement of detectability due to inhomogeneity. The explicit solution for the eigenvector-component distribution is a notable technical contribution that goes beyond eigenvalue-only analyses and could be useful for downstream inference tasks. The non-monotonicity result is falsifiable and directly testable via finite-N numerics.

major comments (2)
  1. [§3 (derivation of the self-consistent equations)] The central claim that the self-consistent equations close exactly for truncated power-law variances (and yield a non-monotonic BBP line) rests on uniform integrability of the requisite moments in the cavity or resolvent fixed-point map. The manuscript should add an explicit verification that the truncation bounds guarantee the necessary moment conditions, especially since power-law tails can produce heavy-tailed effective fields even after truncation.
  2. [§5 (solution for truncated power-law)] The non-monotonicity of the BBP transition is illustrated for specific truncation parameters and power-law exponents. It is unclear whether this feature persists for all admissible truncation bounds or whether it is an artifact of the particular parameter regime chosen to solve the equations; a brief sensitivity analysis with respect to the truncation bounds would strengthen the result.
minor comments (2)
  1. Notation for the variance profile and the truncation parameters should be introduced once and used consistently; currently the same symbol appears to be overloaded in the abstract and the main text.
  2. The manuscript would benefit from a short numerical section comparing the predicted outlier eigenvalue and eigenvector overlap against finite-N simulations for at least two values of the power-law exponent.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading of the manuscript and for the constructive comments, which have helped us improve the rigor and clarity of the presentation. We address each major comment below and have incorporated revisions accordingly.

read point-by-point responses

  1. Referee: [§3 (derivation of the self-consistent equations)] The central claim that the self-consistent equations close exactly for truncated power-law variances (and yield a non-monotonic BBP line) rests on uniform integrability of the requisite moments in the cavity or resolvent fixed-point map. The manuscript should add an explicit verification that the truncation bounds guarantee the necessary moment conditions, especially since power-law tails can produce heavy-tailed effective fields even after truncation.

    Authors: We agree that an explicit check of the moment conditions strengthens the derivation. In the revised manuscript we have added a dedicated paragraph in §3 together with a short appendix (Appendix A) that verifies uniform integrability. For the truncated power-law with finite lower and upper cutoffs, we bound the fourth-moment integral of the effective field and show that the truncation guarantees the requisite integrability for the cavity map to close exactly; the same bounds also control the tails of the resolvent fixed-point iteration. revision: yes

  2. Referee: [§5 (solution for truncated power-law)] The non-monotonicity of the BBP transition is illustrated for specific truncation parameters and power-law exponents. It is unclear whether this feature persists for all admissible truncation bounds or whether it is an artifact of the particular parameter regime chosen to solve the equations; a brief sensitivity analysis with respect to the truncation bounds would strengthen the result.

    Authors: We appreciate the suggestion. We have solved the self-consistent equations numerically for a broader set of truncation bounds (varying both the lower cutoff and the upper cutoff over an order of magnitude) and for several admissible exponents. The non-monotonic shape of the BBP line is preserved in all cases examined. A brief sensitivity analysis, together with two additional panels, has been inserted into §5 of the revised manuscript. revision: yes

Circularity Check

0 steps flagged

Derivation from standard resolvent/cavity methods is self-contained

full rationale

The paper derives exact self-consistent equations for the spectral edges, outlier eigenvalue, and eigenvector component distribution in the high-dimensional limit of the inhomogeneous spiked Wigner model. These follow from standard resolvent or cavity techniques applied to the rank-one perturbation with i.i.d. variance profile drawn from a truncated power-law; the equations are solved numerically for that distribution to obtain the non-monotonic BBP line. No step reduces by construction to its inputs, no load-bearing self-citation chain is present, and the central claims (gapped regime estimator and non-monotonicity) are independent mathematical consequences rather than tautologies or fitted renamings. The derivation is externally falsifiable via finite-N simulations and does not rely on uniqueness theorems imported from the authors' prior work.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The work rests on the standard high-dimensional limit of random matrix theory and the assumption that variances are i.i.d. draws from a truncated power-law; no new entities are postulated.

free parameters (1)
  • power-law exponent and truncation bounds
    Parameters of the truncated power-law distribution used to generate the noise variances; chosen to demonstrate the non-monotonic effect.
axioms (2)
  • domain assumption High-dimensional limit N to infinity with fixed aspect ratio
    Required for the exact equations of spectral edges and eigenvector distribution to hold.
  • domain assumption Rank-one spike perturbation of Wigner matrix
    The model is defined as inhomogeneous rank-one spiked Wigner.

pith-pipeline@v0.9.0 · 5452 in / 1440 out tokens · 42197 ms · 2026-05-10T02:49:48.161605+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

48 extracted references · 27 canonical work pages · 1 internal anchor

  1. [1]

    Miolane, Phase transitions in spiked matrix estimation: information-theoretic analysis (2019), 1806.04343

    L. Miolane, Phase transitions in spiked matrix estimation: information-theoretic analysis (2019), 1806.04343

    work page arXiv 2019

  2. [2]

    I. M. Johnstone, On the distribution of the largest eigenvalue in principal components analysis, Annals of Statistics 29(2), 295 (2001), doi:10.1214/aos/1009210544

    work page doi:10.1214/aos/1009210544 2001

  3. [3]

    Abbe, Community detection and stochastic block models: recent developments, Journal of Machine Learning Research 18(177), 1 (2018)

    E. Abbe, Community detection and stochastic block models: recent developments, Journal of Machine Learning Research 18(177), 1 (2018)

    2018

  4. [4]

    Perry, A

    A. Perry, A. S. Wein, A. S. Bandeira and A. Moitra, Message-passing algorithms for synchronization problems over compact groups, Communications on Pure and Applied Mathematics 71(11), 2275 (2018)

    2018

  5. [5]

    Péché, The largest eigenvalue of small rank perturbations of hermitian random matrices, Probability Theory and Related Fields 64(6), 575 (2006), doi:10.1007/s00440-005-0462-3

    S. Péché, The largest eigenvalue of small rank perturbations of hermitian random matrices, Probability Theory and Related Fields 64(6), 575 (2006), doi:10.1007/s00440-005-0462-3

    work page doi:10.1007/s00440-005-0462-3 2006

  6. [6]

    Tao, Outliers in the spectrum of iid matrices with bounded rank perturbations, Probability Theory and Related Fields 19(6), 721 (2013), doi:10.1007/s00440-011-0397-x

    T. Tao, Outliers in the spectrum of iid matrices with bounded rank perturbations, Probability Theory and Related Fields 19(6), 721 (2013), doi:10.1007/s00440-011-0397-x

    work page doi:10.1007/s00440-011-0397-x 2013

  7. [7]

    J. Baik, G. Ben Arous and S. P\'ech\'e, Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices, Annals of Probability 37(5), 1643 (2005), doi:10.1214/009117905000000233

    work page doi:10.1214/009117905000000233 2005

  8. [8]

    Barbier, T

    J. Barbier, T. Hou, M. Mondelli and M. S \'a enz, The price of ignorance: how much does it cost to forget noise structure in low-rank matrix estimation?, Advances in Neural Information Processing Systems 35, 36733 (2022)

    2022

  9. [9]

    Barbier, F

    J. Barbier, F. Camilli, Y. Xu and M. Mondelli, Information limits and thouless-anderson-palmer equations for spiked matrix models with structured noise, Physical Review Research 7(1), 013081 (2025)

    2025

  10. [10]

    M. Ma \" da, Large deviations for the largest eigenvalue of rank one deformations of Gaussian ensembles , Electronic Journal of Probability 12(none), 1131 (2007), doi:10.1214/EJP.v12-438

    work page doi:10.1214/ejp.v12-438 2007

  11. [11]

    F \'e ral and S

    D. F \'e ral and S. P \'e ch \'e , The largest eigenvalue of rank one deformation of large wigner matrices, Communications in mathematical physics 272(1), 185 (2007)

    2007

  12. [12]

    Knowles and J

    A. Knowles and J. Yin, The isotropic semicircle law and deformation of wigner matrices, Communications on Pure and Applied Mathematics 66(11), 1663 (2013)

    2013

  13. [13]

    E. P. Wigner, On the statistical distribution of the widths and spacings of nuclear resonance levels, In Mathematical Proceedings of the Cambridge Philosophical Society, vol. 47, pp. 790--798. Cambridge University Press (1951)

    1951

  14. [14]

    Benaych-Georges and R

    F. Benaych-Georges and R. R. Nadakuditi, The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices, Advances in Mathematics 227(1), 494 (2011)

    2011

  15. [15]

    E. J. Cand \`e s, J. Romberg and T. Tao, Robust uncertainty principles: Exact signal reconstruction from highly incomplete frequency information, IEEE Transactions on information theory 52(2), 489 (2006)

    2006

  16. [16]

    Deshpande and A

    Y. Deshpande and A. Montanari, Information-theoretically optimal sparse pca, 2014 IEEE International Symposium on Information Theory pp. 2197--2201 (2014)

    2014

  17. [17]

    Lesieur, F

    T. Lesieur, F. Krzakala and L. Zdeborov\'a, Phase transitions in sparse pca, In 2015 IEEE International Symposium on Information Theory (ISIT), pp. 1635--1639, doi:10.1109/ISIT.2015.7282733 (2015)

    work page doi:10.1109/isit.2015.7282733 2015

  18. [18]

    Aubin, B

    B. Aubin, B. Loureiro, A. Maillard, F. Krzakala and L. Zdeborov \'a , The spiked matrix model with generative priors, Advances in Neural Information Processing Systems 32 (2019)

    2019

  19. [19]

    KRIVELEVICH and B

    M. KRIVELEVICH and B. SUDAKOV, The largest eigenvalue of sparse random graphs, Combinatorics, Probability and Computing 12(1), 61–72 (2003), doi:10.1017/S0963548302005424

    work page doi:10.1017/s0963548302005424 2003

  20. [20]

    Chung, L

    F. Chung, L. Lu and V. Vu, Spectra of random graphs with given expected degrees, Proceedings of the National Academy of Sciences 100(11), 6313 (2003), doi:10.1073/pnas.0937490100, https://www.pnas.org/doi/pdf/10.1073/pnas.0937490100

    work page doi:10.1073/pnas.0937490100 2003

  21. [21]

    A. L. Price, N. J. Patterson, R. M. Plenge, M. E. Weinblatt, N. A. Shadick and D. Reich, Principal components analysis corrects for stratification in genome-wide association studies, Nature genetics 38(8), 904 (2006)

    2006

  22. [22]

    R. M. Bell and Y. Koren, Lessons from the netflix prize challenge, SIGKDD Explor. Newsl. 9(2), 75–79 (2007), doi:10.1145/1345448.1345465

    work page doi:10.1145/1345448.1345465 2007

  23. [23]

    Barbier, J.and Camilli, M

    F. Barbier, J.and Camilli, M. Mondelli and M. S \'a enz, Fundamental limits in structured principal component analysis and how to reach them, Proceedings of the National Academy of Sciences 120(30), e2302028120 (2023)

    2023

  24. [24]

    Guionnet, J

    A. Guionnet, J. Ko, F. Krzakala and L. Zdeborov\'a, Low rank matrix estimation with inhomogeneous noise, Information and Inference 14(2) (2025), doi:10.1093/imaiai/iaaf010

    work page doi:10.1093/imaiai/iaaf010 2025

  25. [25]

    Adomaityte, G

    U. Adomaityte, G. Sicuro and P. Vivo, Pca recovery thresholds in low-rank matrix inference with sparse noise (2025), 2511.11927

    work page arXiv 2025

  26. [26]

    Barra, P

    A. Barra, P. Contucci, E. Mingione and D. Tantari, Multi-species mean field spin glasses. rigorous results, In Annales Henri Poincar \'e , vol. 16, pp. 691--708. Springer (2015)

    2015

  27. [27]

    Panchenko, The free energy in a multi-species Sherrington–Kirkpatrick model , The Annals of Probability 43(6), 3494 (2015), doi:10.1214/14-AOP967

    D. Panchenko, The free energy in a multi-species Sherrington–Kirkpatrick model , The Annals of Probability 43(6), 3494 (2015), doi:10.1214/14-AOP967

    work page doi:10.1214/14-aop967 2015

  28. [28]

    Javanmard and A

    A. Javanmard and A. Montanari, State evolution for general approximate message passing algorithms, with applications to spatial coupling, Information and Inference: A Journal of the IMA 2(2), 115 (2013), doi:10.1093/imaiai/iat004

    work page doi:10.1093/imaiai/iat004 2013

  29. [29]

    Barbier, M

    J. Barbier, M. Dia, N. Macris, F. Krzakala, T. Lesieur and L. Zdeborov\' a , Mutual information for symmetric rank-one matrix estimation: A proof of the replica formula, In Proceedings of the 30th International Conference on Neural Information Processing Systems, NIPS'16, p. 424–432. Curran Associates Inc., Red Hook, NY, USA, ISBN 9781510838819 (2016)

    2016

  30. [30]

    F. L. Metz and J. D. Silva, Spectral density of dense random networks and the breakdown of the wigner semicircle law, Phys. Rev. Res. 2, 043116 (2020), doi:10.1103/PhysRevResearch.2.043116

    work page doi:10.1103/physrevresearch.2.043116 2020

  31. [31]

    Dembo, E

    A. Dembo, E. Lubetzky and Y. Zhang, Empirical spectral distributions of sparse random graphs, In In and Out of Equilibrium 3: Celebrating Vladas Sidoravicius, pp. 319--345. Springer (2021)

    2021

  32. [32]

    J. D. Silva and F. L. Metz, Analytic solution of the resolvent equations for heterogeneous random graphs: spectral and localization properties, Journal of Physics: Complexity 3(4), 045012 (2022), doi:10.1088/2632-072X/aca9b1

    work page doi:10.1088/2632-072x/aca9b1 2022

  33. [33]

    D. C. Hoyle and M. Rattray, Statistical mechanics of learning multiple orthogonal signals: asymptotic theory and fluctuation effects, Physical Review E—Statistical, Nonlinear, and Soft Matter Physics 75(1), 016101 (2007)

    2007

  34. [34]

    Paul, Asymptotics of sample eigenstructure for a large dimensional spiked covariance model, Statistica Sinica p

    D. Paul, Asymptotics of sample eigenstructure for a large dimensional spiked covariance model, Statistica Sinica p. 1617 (2007)

    2007

  35. [35]

    Bun, J.-P

    J. Bun, J.-P. Bouchaud and M. Potters, Cleaning large correlation matrices: tools from random matrix theory, Physics Reports 666, 1 (2017)

    2017

  36. [36]

    Perry, A

    A. Perry, A. S. Wein, A. S. Bandeira and A. Moitra, Optimality and sub-optimality of pca i: Spiked random matrix models, Annals of Statistics 46(5), 2416 (2018), doi:10.1214/17-AOS1625

    work page doi:10.1214/17-aos1625 2018

  37. [37]

    O'Rourke and D

    S. O'Rourke and D. Renfrew, Low rank perturbations of large elliptic random matrices, Electronic Journal of Probability 24(3), 1 (2019), doi:10.1214/17-EJP115

    work page doi:10.1214/17-ejp115 2019

  38. [38]

    S. F. Edwards and R. C. Jones, The eigenvalue spectrum of a large symmetric random matrix, Journal of Physics A: Mathematical and General 9(10), 1595 (1976), doi:10.1088/0305-4470/9/10/011

    work page doi:10.1088/0305-4470/9/10/011 1976

  39. [39]

    A. D. Mirlin, Statistics of energy levels and eigenfunctions in disordered systems, Physics Reports 326(5-6), 259 (2000)

    2000

  40. [40]

    F. L. Metz, I. Neri and T. Rogers, Spectral theory of sparse non-hermitian random matrices, Journal of Physics A: Mathematical and Theoretical 52(43), 434003 (2019), doi:10.1088/1751-8121/ab3e63

    work page doi:10.1088/1751-8121/ab3e63 2019

  41. [41]

    Neri and F

    I. Neri and F. L. Metz, Eigenvalue outliers of non-hermitian random matrices with a local tree structure, Physical Review Letters 117(22) (2016), doi:10.1103/physrevlett.117.224101

    work page doi:10.1103/physrevlett.117.224101 2016

  42. [42]

    Neri and F

    I. Neri and F. L. Metz, Linear stability analysis of large dynamical systems on random directed graphs, Phys. Rev. Res. 2, 033313 (2020), doi:10.1103/PhysRevResearch.2.033313

    work page doi:10.1103/physrevresearch.2.033313 2020

  43. [43]

    F. L. Metz and I. Neri, Localization and universality of eigenvectors in directed random graphs, Phys. Rev. Lett. 126, 040604 (2021), doi:10.1103/PhysRevLett.126.040604

    work page doi:10.1103/physrevlett.126.040604 2021

  44. [44]

    Cizeau and J.-P

    P. Cizeau and J.-P. Bouchaud, Theory of l \'e vy matrices , Physical Review E 50(3), 1810 (1994)

    1994

  45. [45]

    Tapias and P

    D. Tapias and P. Sollich, Multifractality and statistical localization in highly heterogeneous random networks, Europhysics Letters 144(4), 41001 (2023), doi:10.1209/0295-5075/ad1001

    work page doi:10.1209/0295-5075/ad1001 2023

  46. [46]

    J. D. da Silva, D. Tapias, P. Sollich and F. L. Metz, Spectral properties, localization transition and multifractal eigenvectors of the Laplacian on heterogeneous networks , SciPost Phys. 18, 047 (2025), doi:10.21468/SciPostPhys.18.2.047

    work page doi:10.21468/scipostphys.18.2.047 2025

  47. [47]

    Bousseyroux and M

    P. Bousseyroux and M. Potters, The eigenvalues and eigenvectors of finite-rank normal perturbations of large rotationally invariant non-hermitian matrices, arXiv preprint arXiv:2601.10427 (2026)

    work page internal anchor Pith review arXiv 2026

  48. [48]

    Feinberg and A

    J. Feinberg and A. Zee, Non-hermitian random matrix theory: Method of hermitian reduction, Nuclear Physics B 504(3), 579 (1997), doi:10.1016/S0550-3213(97)00502-6

    work page doi:10.1016/s0550-3213(97)00502-6 1997