Pith. sign in

REVIEW 4 major objections 5 minor 46 references

An SIS epidemic on a complex network can be collapsed to a single scalar differential equation on a spectral submanifold—the smoothest invariant curve through the disease-free equilibrium—that still reproduces every node's infection traject

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-07-31 22:57 UTC pith:E5Y66Z6H

load-bearing objection A solid new application of SSM to network dynamics, but the unproved heteroclinic-capture assumption and the missing mean-field baseline keep the core claim from standing without revision. the 4 major comments →

arxiv 2607.24121 v1 pith:E5Y66Z6H submitted 2026-07-27 physics.soc-ph math.DSq-bio.QM

Nonlinear Model Reduction of Complex Networks via Spectral Submanifolds

Kaviya Bhaskaran , Shobhit Jain , Mingwu Li This is my paper
classification physics.soc-ph math.DSq-bio.QM
keywords Spectral submanifoldsSSM reductionnode-level model reductioncomplex networkstipping-point predictionSusceptible-Infected-Susceptible modelhigher-order interactionsPadé globalization
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to show that complicated networks do not always need N-dimensional simulation: by constructing a spectral submanifold attached to the slowest-decaying mode at the disease-free equilibrium, the full SIS epidemic reduces to one scalar equation plus a lifting map that rebuilds each node's infection level. If that works, one short differential equation and a handful of coefficients replace a large coupled system while retaining the information that matters—who gets infected, when, and how large the endemic state becomes. The strongest claim is that this one-dimensional reduction predicts the epidemic threshold even at quadratic order, with higher-order terms or a rational extension (gSSM) recovering post-onset saturation, and that it consistently beats classical spectral and mean-field reductions across homogeneous, heterogeneous, modular, empirical, and higher-order-interaction networks. A sympathetic reader would care because it promises interpretable, node-resolved forecasting in epidemiology and ecology without discarding the nonlinearity that creates tipping points.

Core claim

The paper establishes that, for SIS epidemics on networks, the nonlinear dynamics are governed after a short transient by a single latent coordinate on the one-dimensional spectral submanifold tangent to the leading eigenvector of the linearized disease-free equilibrium. Solving the invariance equation order-by-order yields both the scalar reduced dynamics and the lifting map that reconstructs the full N-dimensional state; where the Taylor series converges poorly, a Padé-type rational continuation (gSSM) extends the reduced model to the high-prevalence regime. The reduced system reproduces the mean prevalence, node-level trajectories, community-level curves, and the tipping onset on Erdős–Ré

What carries the argument

The central object is the spectral submanifold (SSM): the smoothest invariant manifold tangent to the spectral subspace of the linearized network at the selected equilibrium, here the one-dimensional subspace of the slowest-decaying mode. Its reduced dynamics are generated by the invariance equation, DW(η)R(η) = f(W(η)), solved with Taylor expansions to produce the scalar drift R(η) and the lifting coefficients W(η) that rebuild the network state. The globalized extension (gSSM) replaces the truncated Taylor drift with a Padé-type rational function, extending the reduced model beyond the local convergence radius—the step that recovers post-onset saturation on hub-localized networks. The mach

Load-bearing premise

The reduction is built only at the healthy equilibrium, and the whole argument rests on the stated but unproved premise that the one-dimensional curve found there continues all the way to the infected steady state, carrying the transition between them; if that connection does not exist for some network, the reduced equation can only describe the early near-healthy phase.

What would settle it

Within the regime the paper assumes (a real, spectrally separated leading eigenvalue), take a network that satisfies that condition but whose endemic equilibrium lies away from the manifold extending the leading eigenvector—detectable by comparing full-system steady states with the O(20) and gSSM lifted steady states across a fine β/γ grid. If, for any β/γ above the threshold, the reduced scalar equation lacks a positive root inside the Padé convergence domain or the lifted state disagrees with the full endemic state, the central claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • SIS epidemics on homogeneous and small-world networks are effectively one-dimensional after a short transient; a root of the reduced scalar drift gives the endemic steady state, so final prevalence can be computed from a polynomial.
  • The epidemic threshold is captured even by the quadratic truncation, meaning a low-order SSM computation can estimate the onset of sustained activity without a full network simulation.
  • Node-level forecasts are available from the same scalar coordinate via the lifting map, giving per-node infection curves that mean-field and classical spectral projections do not provide.
  • On scale-free networks, modular-bottleneck networks, and higher-order triadic models, low-order truncations are biased; higher-order Taylor terms or gSSM restore agreement with the full system.
  • The same workflow extends beyond SIS to generalized Lotka–Volterra, gene-regulatory, and logistic–diffusion networked equations, each collapsed to a one-dimensional reduced model, with the logistic–diffusion case exact at quadratic order.

Where Pith is reading between the lines

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

  • The reduction's reach depends on a single slow curve persisting all the way from the healthy state to the endemic state; the most natural stress test is a bistable higher-order regime with a saddle-node where no single heteroclinic connection exists, which would require a two-branch or higher-dimensional manifold construction.
  • The convergence pattern suggests a practical pre-analysis: compute the inverse participation ratio of the leading eigenvector to predict whether low-order or high-order/gSSM reduction will be needed, since hub localization marks the regime where polynomial truncations shrink.
  • Because the reduced scalar equation is polynomial or rational, it may yield closed-form expressions for effective reproduction numbers and final-size relations in heterogeneous networks, connecting this reduction to standard epidemic theory.
  • A data-driven version that learns the SSM from node-level time series would make the reduction usable when the network topology is unknown; the paper cites such methods but does not develop that step here.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a model-reduction framework for nonlinear network dynamics based on spectral submanifolds (SSMs) and a Padé-based globalization (gSSM). For SIS epidemic dynamics on synthetic and empirical networks, the authors construct a one-dimensional invariant manifold tangent to the dominant eigenvector at the disease-free equilibrium, derive the reduced dynamics and lifting map via the parameterization method, and test the reduced model's ability to reproduce node-level trajectories, mean prevalence, and steady-state response curves across β/γ sweeps. The paper also extends the approach to higher-order triadic SIS dynamics and to Lotka–Volterra, gene-regulatory, and logistic-diffusion models, reporting that SSM/gSSM consistently outperform spectral and mean-field baselines.

Significance. If the central claims hold, the paper offers a notable contribution: a single scalar ODE plus a lifting map that reconstructs full node-level behavior in N-dimensional epidemic networks, with a rigorous invariance-equation derivation. The homological coefficient recursions in Appendix A are explicit and standard, the reduced coefficients are derived from the model rather than fitted to full-system output, and the authors provide code and reproducibility documentation. The reported agreement at moderate-to-high truncation orders in homogeneous networks, and the improved performance of gSSM on heterogeneous networks, are plausible and potentially useful. However, the broad 'consistently outperform' claim is currently not fully supported because no mean-field baseline is implemented, only one network instance per topology is used, and the post-onset predictions rest on an unproved assertion about global heteroclinic capture by the local one-dimensional SSM.

major comments (4)
  1. [Sec. II.B and Figs. 2–5] The load-bearing assumption that the one-dimensional SSM constructed at the disease-free equilibrium contains the endemic equilibrium and the heteroclinic orbit connecting them is asserted without proof. The text states this directly in Sec. II.B: 'the resulting one-dimensional SSM recovers the heteroclinic orbit connecting to the other fixed point.' This is not a consequence of local SSM existence/uniqueness theory, and for β/γ above threshold the disease-free equilibrium is unstable, so the relevant object is an unstable manifold. The root analysis in Fig. 2(d,h,l) only locates a positive real zero of the truncated reduced vector field; it never checks whether the lifting map evaluated at that zero reproduces the full endemic equilibrium x*. I ask the authors to verify, for each network and parameter value, that ||W(ρ*) − x*|| is small, and to state precisely in what sense the heterocl
  2. [Abstract, Sec. III, Fig. 4] The abstract and conclusion claim that SSM/gSSM 'consistently outperform classical spectral and mean-field methods,' but no mean-field baseline is implemented or plotted anywhere in the manuscript or SM. The sweeps in Fig. 4 compare the full system, spectral reduction, modified spectral reduction, and SSM/gSSM only. The conclusion repeats the mean-field claim without supporting data. Since this is a central advertised result, the authors must either add the relevant mean-field baselines (e.g., heterogeneous or quenched mean-field) across the same synthetic and empirical networks, or explicitly qualify the claim.
  3. [Sec. III (comparison protocol)] The comparison is based on a single network instance for each topology. The text states 'we use the same network instance for each synthetic topology—ER, SW, SF—and the same empirical contact networks... across all analyses,' and Table S1 confirms 'Fixed per topology, reused in all tests.' With N=200 and stochastic network models, especially SF and modular SBM, individual realizations can be atypical. The 'consistently outperform' claim would be much better supported by ensemble statistics (e.g., median and spread over at least a few dozen realizations per topology, with the same SSM order). As written, the reader cannot assess whether the reported accuracy and ranking are robust or instance-specific.
  4. [Sec. III, Fig. 4(l) and Sec. II.B] On the Rural network, even high-order Taylor SSM reductions underestimate the steady state for β/γ > 0.2, and accuracy is restored only by gSSM. On SF, the positive real root of the truncated reduced dynamics lies near the estimated convergence boundary (Fig. 2l). This is a concrete regime where the Taylor-based SSM alone does not support the claimed superiority, and the gSSM results are presented without a quantitative error metric or convergence study. Please report the gSSM error (e.g., MSE or steady-state error vs. full system) for these cases and clarify the criterion for 'where necessary' in selecting gSSM.
minor comments (5)
  1. [Sec. III.B] Typo: 'Rural, srepresenting' should be 'Rural, representing'.
  2. [Sec. III.B, Fig. 4(l) caption/description] Typo: 'mises-timates' should be 'misestimates'.
  3. [Sec. III, Fig. 2 panel references] The text says 'Fig. 2d,h,j' but the SF panel is labeled (l); please correct to (d,h,l).
  4. [Table S1 vs. Sec. III] The main text fixes η0=0.01, while Table S1 lists 'η0 ∈ [0.01,0.05]'; please reconcile or specify when other values are used.
  5. [Appendix B] The Padé construction is described as data-driven via [21], but the paper is equation-driven. Clarify whether the Padé coefficients are determined purely from the Taylor coefficients of the reduced dynamics or involve any trajectory information.

Circularity Check

0 steps flagged

No significant circularity: the SSM/gSSM reduction is derived from the model equations via the invariance equation, with no fitted parameter renamed as a prediction and no load-bearing self-citation chain.

full rationale

The paper's central pipeline is equation-driven: the SSM pair (W,R) is defined by the invariance equation DW(η)R(η)=f(W(η)) (Eq. 2), and the Taylor coefficients of W and R are solved order-by-order from the full network vector field via linear homological equations (Appendix A, Eqs. A5-A10). No parameter is fitted to full-system trajectories or to the reported observables; the lifting map and reduced dynamics are determined entirely by the model itself. The comparisons against full-order SIS simulations (Figs. 2-5) are therefore genuine benchmarks of a derived reduction, not circular predictions. The cited existence/uniqueness theory [22] and parameterization method [23] are established, externally published results, and the present paper does not use its own conclusions to justify them; SSMTool [31] is a software implementation rather than a load-bearing self-citation. The Padé-based gSSM is likewise a prior method [21] applied here without refitting to the full model. Two concerns are real but are correctness/convergence issues, not circularity: (i) the asserted heteroclinic capture, 'In all examples considered in this paper... the resulting one-dimensional SSM recovers the heteroclinic orbit connecting to the other fixed point' (Sec. II.B), is unproved; and (ii) the root criterion defines a 'non-spurious' root by persistence under increasing truncation order, which is self-referential as a heuristic. Neither reduces a claimed prediction to an input: the endemic equilibrium is not used in constructing the SSM, and the truncated root analysis is still derived from the model's own coefficients. Hence no prediction is equivalent to its inputs by construction.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

No new physical entities are introduced; SSM/gSSM are mathematical constructs. The main imported content is SSM existence theory and SSMTool software from the same research lineage (Jain, Li, Haller). The method has no fitted parameters in the strict sense, but several hand-chosen algorithmic settings (truncation order, Pade type, initialization amplitude, root-radius heuristic) control the reported accuracy.

free parameters (4)
  • Taylor truncation order p = 20 (max); results depend on p (O(2)..O(20))
    Central accuracy claims are conditioned on the truncation order; there is no a priori criterion for choosing p, and on SF/Rural networks only high p or gSSM works.
  • Pade approximant type [8/7] = [8/7] built from order-15 Taylor
    Chosen by hand as 'effective' across studied models; controls gSSM behavior and the claimed global recovery.
  • Initial reduced coordinate η0 = 0.01
    Small prescribed amplitude chosen to start near the equilibrium; initialization protocol affects transient comparison.
  • Empirical Taylor radius R_P = computed from outer half of highest-order root moduli
    Used to classify roots as spurious vs. bona fide; heuristic, no cited formal basis.
axioms (5)
  • standard math Under standard nonresonance and spectral-quotient conditions, a unique C^r SSM tangent to the selected spectral subspace exists at the origin (Haller–Ponsioen 2016).
    Invoked at Sec. II B; conditions are not verified numerically in the paper.
  • ad hoc to paper The one-dimensional SSM constructed locally at the disease-free equilibrium extends to contain the endemic equilibrium and the heteroclinic orbit connecting them.
    Stated without proof in Sec. II B ('In all examples considered... the resulting one-dimensional SSM recovers the heteroclinic orbit...'); this is load-bearing for post-onset predictions.
  • domain assumption Trajectories of interest are attracted to the SSM; transient fast components decay and the 1D slow dynamics dominate after a short transient.
    Used throughout; explicitly acknowledged in SM S1 that fast transient is absent from the reduced model.
  • standard math Pade approximants of type [8/7] converge to the global reduced vector field on the relevant interval (Montessus-type convergence).
    Appendix B; convergence theorems for meromorphic functions are cited, but pole locations for the specific networks are not analyzed.
  • domain assumption For HOI SIS, simplicial triangle closure is a valid representation of higher-order contagion.
    Appendix S5; standard in HOI literature but a modeling choice, not derived.

pith-pipeline@v1.3.0-alltime-deepseek · 26324 in / 13746 out tokens · 115140 ms · 2026-07-31T22:57:11.773842+00:00 · methodology

0 comments
read the original abstract

Complex networked systems are prevalent in biology, engineering, and the social sciences, yet their high-dimensional, nonlinear dynamics pose major challenges for analysis and prediction. A mathematically rigorous route to simplification is to represent system behavior on a low-dimensional, smooth invariant manifold known as a spectral submanifold (SSM). Here we present a comprehensive SSM reduction framework and its globalized extension (gSSM) for dimensionality reduction in large-scale nonlinear networks. Our approach yields accurate global and node-level predictions across synthetic and real networks, including highly heterogeneous topologies and systems with higher-order interactions. Crucially, SSM is a robust tipping-point predictor: even at low truncation order (e.g., $O(2)$) it reliably identifies the onset of sustained activity, while higher orders and gSSM capture post-onset amplitudes and saturation. Consistently, the reduction collapses the full network dynamics to a one-dimensional system, offering clarity and efficiency. Across all the realizations, SSM/gSSM consistently outperform classical spectral and mean-field methods in modeling critical transitions at both microscopic and macroscopic scales, establishing SSM-based reduction as a robust, interpretable tool for nonlinear networked systems with broad applicability to epidemiology, ecology, and engineered networks.

Figures

Figures reproduced from arXiv: 2607.24121 by Kaviya Bhaskaran, Mingwu Li, Shobhit Jain.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic of the SSM reduction pipeline for complex networks. Starting from node dynamics and topology, a spectral [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. SIS dynamics and comparison of full and reduced trajectories Erd˝os–R´enyi (ER), Scale-Free (SF), and Small-World [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Modular-bottleneck benchmark using a two-block [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. One-parameter sweeps of final mean infection [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Final mean infection [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

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

46 extracted references · 1 linked inside Pith

  1. [1]

    Pascual and J

    M. Pascual and J. A. Dunne, eds.,Ecological Networks: Linking Structure to Dynamics in Food Webs, Santa Fe Institute Studies on the Sciences of Complexity (Oxford University Press, Oxford, 2006)

    2006

  2. [2]

    R. M. Anderson and R. M. May,Infectious Diseases of Humans: Dynamics and Control, Oxford Science Publications (Oxford University Press, Oxford, 1991)

    1991

  3. [3]

    Dobson, B

    I. Dobson, B. A. Carreras, V. E. Lynch, and D. E. Newman, Complex systems analysis of series of blackouts: Cascading failure, critical points, and self-organization, Chaos17, 026103 (2007)

    2007

  4. [4]

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams, Learning representations by back-propagating errors, Nature323, 533 (1986)

    1986

  5. [5]

    J. Gao, B. Barzel, and A.-L. Barab´ asi, Universal resilience patterns in complex networks, Nature530, 307 (2016)

    2016

  6. [6]

    Landi, H

    P. Landi, H. O. Minoarivelo, ˚A. Br¨ annstr¨ om, C. Hui, and U. Dieckmann, Complexity and stability of ecological networks: a review of the theory, Population Ecology60, 319 (2018)

    2018

  7. [7]

    Jiang, Z.-G

    J. Jiang, Z.-G. Huang, T. P. Seager, A. Hastings, and Y.-C. Lai, Predicting tipping points in mutualistic networks through dimension reduction, Proceedings of the National Academy of Sciences115, E639 (2018)

    2018

  8. [8]

    Laurence, N

    E. Laurence, N. Doyon, L. J. Dub´ e, and P. Desrosiers, Spectral dimension reduction of complex dynamical networks, Physical Review X9, 011042 (2019)

    2019

  9. [9]

    Masuda and P

    N. Masuda and P. Kundu, Dimension reduction of dynamical systems on networks with leading and nonleading eigenvectors of adjacency matrices, Physical Review Research4, 023257 (2022)

    2022

  10. [10]

    Vegu´ e, V

    M. Vegu´ e, V. Thibeault, P. Desrosiers, and A. Allard, Dimension reduction of dynamics on modular and heterogeneous directed networks, PNAS Nexus2, pgad150 (2023)

    2023

  11. [11]

    C. Tu, P. D’Odorico, and S. Suweis, Dimensionality reduction of complex dynamical systems, iScience24, 101912 (2021)

    2021

  12. [12]

    C. Tu, Y. Wu, J. Luo, Y. Jiang, and X. Pan, Dimensionality reduction in discrete-time dynamical systems, Communications in Nonlinear Science and Numerical Simulation123, 107268 (2023)

    2023

  13. [13]

    C. Tu, J. Luo, Y. Fan, and X. Pan, Dimensionality reduction in stochastic complex dynamical networks, Chaos, Solitons & Fractals175, 114034 (2023)

    2023

  14. [14]

    Thibeault, G

    V. Thibeault, G. St-Onge, L. J. Dub´ e, and P. Desrosiers, Threefold way to the dimension reduction of dynamics on networks: An application to synchronization, Physical Review Research2, 043215 (2020)

    2020

  15. [15]

    C. Wu, D. Duan, and R. Xiao, A novel dimension reduction method with information entropy to evaluate network resilience, Physica A: Statistical Mechanics and its Applications620, 128727 (2023)

    2023

  16. [16]

    F. G. Mohammadi, M. H. Amini, and H. R. Arabnia, Applications of nature-inspired algorithms for dimension reduction: enabling efficient data analytics, inOptimization, learning, and control for interdependent complex networks(Springer,

  17. [17]

    Prasse and P

    B. Prasse and P. V. Mieghem, Predicting network dynamics without requiring the knowledge of the interaction graph, Proceedings of the National Academy of Sciences119, e2205517119 (2022)

    2022

  18. [18]

    Prasse, K

    B. Prasse, K. Devriendt, and P. V. Mieghem, Clustering for epidemics on networks: A geometric approach, Chaos31, 063115 (2021)

    2021

  19. [19]

    Y. Ding, Z. Huang, M. Magdon-Ismail, and J. Gao, Predicting time series of networked dynamical systems without knowing topology, arXiv preprint arXiv:2412.18734 (2024)

    Pith/arXiv arXiv 2024

  20. [20]

    Thibeault, A

    V. Thibeault, A. Allard, and P. Desrosiers, The low-rank hypothesis of complex systems, Nature Physics20, 294 (2024)

    2024

  21. [21]

    Kasz´ as and G

    B. Kasz´ as and G. Haller, Globalizing manifold-based reduced models for equations and data, Nature Communications16, 61252 (2025)

    2025

  22. [22]

    Haller and S

    G. Haller and S. Ponsioen, Nonlinear normal modes and spectral submanifolds: existence, uniqueness and use in model reduction, Nonlinear Dynamics86, 1493 (2016)

    2016

  23. [23]

    Jain and G

    S. Jain and G. Haller, How to compute invariant manifolds and their reduced dynamics in high-dimensional finite element models, Nonlinear Dynamics107, 1417 (2022)

    2022

  24. [24]

    Breunung and G

    T. Breunung and G. Haller, Explicit backbone curves from spectral submanifolds of forced-damped nonlinear mechanical systems, Proceedings of the Royal Society A474, 20180083 (2018)

    2018

  25. [25]

    Ponsioen, L

    S. Ponsioen, L. Renson, G. W. H. van der Veen, and G. Haller, Model reduction to spectral submanifolds and forced response curve computation, Journal of Sound and Vibration488, 115640 (2020)

    2020

  26. [26]

    M. Li, S. Jain, and G. Haller, Nonlinear analysis of forced mechanical systems with internal resonance using spectral submanifolds, part i: Periodic response and forced response curve, Nonlinear Dynamics110, 1005 (2022)

    2022

  27. [27]

    Li and G

    M. Li and G. Haller, Nonlinear analysis of forced mechanical systems with internal resonance using spectral submanifolds, part ii: Bifurcation and quasi-periodic response, Nonlinear Dynamics110, 1045 (2022)

    2022

  28. [28]

    Cenedese, J

    M. Cenedese, J. Ax ˚ as, B. B¨ auerlein, K. Avila, and G. Haller, Data-driven modeling and prediction of non-linearizable dynamics via spectral submanifolds, Nature Communications13, 872 (2022)

    2022

  29. [29]

    A. Liu, J. Ax ˚ as, and G. Haller, Data-driven modeling and forecasting of chaotic dynamics on inertial manifolds constructed as spectral submanifolds, Chaos34, 033140 (2024)

    2024

  30. [30]

    nonlinear spectral model reduction of networked systems

    K. Bhaskaran, S. Jain, and M. Li, Supplemental material for “nonlinear spectral model reduction of networked systems” (2025), contains extended derivations, implementation details, and additional figures

    2025

  31. [31]

    S. Jain, M. Li, T. Thurnher, and G. Haller, Ssmtool 2.6: Computation of invariant manifolds in high-dimensional mechanics problems (v2.6), Zenodo (2024), software

    2024

  32. [32]

    Ponsioen, T

    S. Ponsioen, T. Pedergnana, and G. Haller, Analytic prediction of isolated forced response curves from spectral submani- 33 folds, Nonlinear Dynamics98, 2755 (2019)

    2019

  33. [33]

    SocioPatterns Collaboration, Hospital ward contact network, lyon, 2010,https://www.sociopatterns.org/datasets/ hospital-ward-dynamic-contact-network/(2010), accessed 2025

    2010

  34. [34]

    SocioPatterns Collaboration, Office building contact network, france, 2013,https://www.sociopatterns.org/datasets/ contacts-in-a-workplace/(2013), accessed 2025

    2013

  35. [35]

    SocioPatterns Collaboration, Rural malawi contact network,https://www.sociopatterns.org/datasets/rural-malawi/ (2014), accessed 2025

    2014

  36. [36]

    Iacopini, G

    I. Iacopini, G. Petri, A. Barrat, and V. Latora, Simplicial models of social contagion, Nature Communications10, 2485 (2019)

    2019

  37. [37]

    Battiston, G

    F. Battiston, G. Cencetti, I. Iacopini, V. Latora, M. Lucas, A. Patania, J.-G. Young, and G. Petri, Networks beyond pairwise interactions: Structure and dynamics, Physics Reports874, 1 (2020)

    2020

  38. [38]

    Battiston, E

    F. Battiston, E. Amico, A. Barrat, G. Bianconi, G. F. de Arruda, B. Franceschiello, I. Iacopini, S. K´ efi, V. Latora, Y. Moreno, M. M. Murray, T. P. Peixoto, F. Vaccarino, and G. Petri, The physics of higher-order interactions in complex systems, Nature Physics17, 1093 (2021)

    2021

  39. [39]

    Szaksz, G

    B. Szaksz, G. Orosz, and G. Stepan, Spectral submanifolds in time delay systems, Nonlinear Dynamics113, 14265 (2025)

    2025

  40. [40]

    H. S. Wall,Analytic Theory of Continued Fractions(D. Van Nostrand, New York, 1948)

    1948

  41. [41]

    G. A. Baker and P. Graves-Morris,Pad´ e Approximants, 2nd ed. (Cambridge University Press, Cambridge, UK, 1996)

    1996

  42. [42]

    A. Cuyt, V. Petersen, B. Verdonk, H. Waadeland, and W. B. Jones,Handbook of Continued Fractions for Special Functions (Springer, New York, 2008)

    2008

  43. [43]

    de Montessus de Ballore, Sur les fractions continues alg´ ebriques, Bulletin de la Soci´ et´ e Math´ ematique de France30, 28 (1902)

    R. de Montessus de Ballore, Sur les fractions continues alg´ ebriques, Bulletin de la Soci´ et´ e Math´ ematique de France30, 28 (1902)

    1902

  44. [44]

    Brezinski,Pad´ e-Type Approximation and General Orthogonal Polynomials(Birkh¨ auser, Basel, 1991)

    C. Brezinski,Pad´ e-Type Approximation and General Orthogonal Polynomials(Birkh¨ auser, Basel, 1991)

    1991

  45. [45]

    P. R. Graves-Morris,Vector and Matrix Pad´ e Approximations(Clarendon Press, Oxford, 1979)

    1979

  46. [46]

    Vanhems, A

    P. Vanhems, A. Barrat, C. Cattuto, J.-F. Pinton, N. Khanafer, C. Regis, B.-a. Kim, B. Comte, and N. Voirin, Estimating potential infection transmission routes in hospital wards using wearable proximity sensors (2013)

    2013