Learning graphons from data: Random walks, transfer operators, and spectral clustering

arxiv: 2507.18147 · v2 · submitted 2025-07-24 · 📊 stat.ML

Learning graphons from data: Random walks, transfer operators, and spectral clustering

Stefan Klus , Jason J. Bramburger This is my paper

Pith reviewed 2026-05-19 03:07 UTC · model grok-4.3

classification 📊 stat.ML

keywords graphonsrandom walkstransfer operatorsspectral clusteringstochastic processesdata-driven reconstructiongraph limits

0 comments p. Extension

The pith

Signals from stochastic processes can be modeled as random walks on graphons to enable spectral clustering and reconstruction of the underlying structure from data.

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

The paper links time series signals that switch between a continuum of values to random walk dynamics on a graphon, the limiting object for sequences of large graphs. It defines associated transfer operators such as the Koopman and Perron-Frobenius operators and shows these can be estimated directly from observed signal values. The eigenvalues and eigenfunctions of the estimated operators then support cluster detection, extending spectral clustering methods beyond finite graphs. When the underlying walk is reversible, the same data-driven estimates also recover the transition probability densities and the graphon itself. Demonstrations on both synthetic examples and real signals such as daily temperatures and stock indices illustrate the approach.

Core claim

Transfer operators for random walks on graphons can be estimated from a single observed signal, their spectral properties detect clusters in the continuum state space, and the operators allow recovery of the transition densities and, for reversible walks, the graphon itself.

What carries the argument

Transfer operators (Koopman and Perron-Frobenius) for random walks on graphons, which map functions on the continuum state space and whose data-driven approximations yield eigenvalues and eigenfunctions for analysis.

If this is right

Spectral clustering extends from finite graphs to signals with values in a continuum without requiring an explicit graph construction.
Transition probability densities between states can be recovered directly from the signal without prior knowledge of the graph structure.
For reversible processes the full graphon can be reconstructed, supplying a compact description of the limiting dynamics.
The same operator estimates apply uniformly to both synthetic benchmarks and real-world time series such as temperature records or financial indices.

Where Pith is reading between the lines

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

The framework could be tested on signals from systems known to have continuous state spaces, such as diffusion processes on manifolds, to check whether graphon approximations capture the observed clustering.
Non-reversible cases might still permit approximate reconstruction by focusing on the forward or backward operator alone rather than requiring symmetry.
Connections to ergodic theory suggest the approach could help identify invariant measures or mixing rates in long time series without discretizing the state space first.

Load-bearing premise

The observed signal is produced by a random walk on some underlying graphon.

What would settle it

A concrete falsifier would be a signal whose empirically estimated transition statistics cannot be matched by any graphon random walk, or where the reconstructed graphon fails to reproduce the observed transition probabilities on held-out segments of the data.

read the original abstract

Many signals evolve in time as a stochastic process, randomly switching between states over discretely sampled time points. Here we make an explicit link between the underlying stochastic process of a signal that can take on a bounded continuum of values and a random walk process on a graphon. Graphons are infinite-dimensional objects that represent the limit of convergent sequences of graphs whose size tends to infinity. We introduce transfer operators, such as the Koopman and Perron--Frobenius operators, associated with random walk processes on graphons and then illustrate how these operators can be estimated from signal data and how their eigenvalues and eigenfunctions can be used for detecting clusters, thereby extending conventional spectral clustering methods from graphs to graphons. Furthermore, we show that it is also possible to reconstruct transition probability densities and, if the random walk process is reversible, the graphon itself using only the signal. The resulting data-driven methods are applied to a variety of synthetic and real-world signals, including daily average temperatures and stock index values.

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.

Desk Editor's Note private letter to a colleague

This paper links continuum stochastic signals to graphon random walks and uses estimated transfer operators for spectral clustering plus reconstruction of transition densities or the graphon itself under reversibility.

read the letter

The punchline is that the authors treat bounded continuum-valued time series as random walks on graphons, then estimate the associated transfer operators from data to extend spectral clustering and reconstruct transition densities or the graphon when the process is reversible. They apply the approach to synthetic cases and real signals like daily temperatures and stock indices. What is new is the explicit construction of those operators for graphon random walks combined with the data-driven reconstruction step, which builds on prior graphon and Koopman work but puts them together for this purpose. The applications section shows the methods can handle practical signals without needing an explicit graph. The soft spot is that the reconstruction claim depends on the modeling assumption that the observed signal really comes from a graphon random walk, and the abstract gives little detail on the estimation procedure or error analysis, so it is hard to judge robustness without the full derivations. No obvious internal contradictions show up from the given material. This is for people working in statistical machine learning or dynamical systems who care about graph limits, spectral methods on infinite objects, or data-driven modeling of continuous-state processes. A reader already familiar with transfer operators or graphons would get the most out of it and see how the extension plays out in examples. It deserves a serious referee because the framing is coherent and the real-data tests are there, even if the theoretical steps will need checking and possibly tightening. I would send it to peer review rather than desk reject.

Referee Report

2 major / 2 minor

Summary. The paper establishes a link between stochastic processes on bounded continuous state spaces and random walks on graphons (as limits of convergent graph sequences). It defines associated transfer operators (Koopman and Perron-Frobenius) on graphons, shows how these can be estimated from discretely sampled signal data, uses their spectral properties to extend spectral clustering to the graphon setting, and claims that transition probability densities (and the graphon itself, under reversibility) can be reconstructed from the observed signal alone. The approach is illustrated on synthetic examples and real signals such as daily temperatures and stock indices.

Significance. If the reconstruction and estimation results hold with the stated guarantees, the work provides a principled data-driven route to infer graphon structure and dynamics directly from time series without constructing explicit graphs. It bridges transfer-operator methods from dynamical systems with graph-limit theory, extending spectral clustering beyond finite graphs in a manner that could apply to large-scale or continuum network models. The absence of free parameters in the core operator constructions and the explicit modeling assumption (random walk on an underlying graphon) are strengths that support falsifiability and reproducibility.

major comments (2)

[Reconstruction result (likely §4 or §5)] The central reconstruction claim for transition densities and the graphon (under reversibility) is load-bearing; the abstract states it follows from the signal, but the manuscript must supply the explicit identifiability argument and any consistency rates for the estimator in the relevant theorem or proposition, as the current presentation leaves open whether discretization or finite-sample effects undermine the infinite-graphon limit.
[Operator estimation section] The estimation procedure for the transfer operators from discrete-time samples must include a consistency analysis with respect to the graphon measure; without it, the extension of spectral clustering to graphons rests on an unverified approximation step that could affect the eigenvalue/eigenfunction recovery used for clustering.

minor comments (2)

[Preliminaries / notation] Clarify notation for the graphon W and the associated measure throughout; ensure the random-walk transition kernel is defined uniformly in both the finite-graph and graphon cases.
[Applications] Add a short discussion of how the reversibility assumption is checked or relaxed in the real-data examples (temperatures, stock indices).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment and the recommendation of minor revision. The comments correctly identify areas where the reconstruction and estimation arguments can be strengthened with additional explicit statements. We address each point below and will revise the manuscript accordingly.

read point-by-point responses

Referee: [Reconstruction result (likely §4 or §5)] The central reconstruction claim for transition densities and the graphon (under reversibility) is load-bearing; the abstract states it follows from the signal, but the manuscript must supply the explicit identifiability argument and any consistency rates for the estimator in the relevant theorem or proposition, as the current presentation leaves open whether discretization or finite-sample effects undermine the infinite-graphon limit.

Authors: We agree that the reconstruction claims require a more explicit identifiability argument. The manuscript derives the transition densities from the estimated Perron–Frobenius operator and recovers the graphon (under reversibility) via the stationary measure and the operator kernel. To close the gap on discretization and finite-sample effects, we will insert a new proposition in §4 that states the identifiability conditions (injectivity of the operator-to-kernel map under the graphon measure) together with consistency rates obtained from empirical-process bounds on the sampled trajectories. These rates will be stated in the infinite-graphon limit as the number of observations tends to infinity and the discretization mesh is refined. revision: yes
Referee: [Operator estimation section] The estimation procedure for the transfer operators from discrete-time samples must include a consistency analysis with respect to the graphon measure; without it, the extension of spectral clustering to graphons rests on an unverified approximation step that could affect the eigenvalue/eigenfunction recovery used for clustering.

Authors: We accept that a formal consistency result is needed. The current estimation approximates the graphon-induced integral operators by Monte-Carlo averages along observed trajectories. We will add a consistency theorem to the operator-estimation section that establishes convergence of the empirical operators to their population counterparts in the Hilbert–Schmidt norm with respect to the graphon measure, under standard ergodicity and moment assumptions on the underlying stochastic process. The theorem will also quantify the effect on the recovered eigenvalues and eigenfunctions, thereby justifying the spectral clustering procedure in the graphon setting. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation applies standard transfer-operator theory to graphons

full rationale

The paper links observed signals to random walks on graphons via transfer operators (Koopman/Perron-Frobenius), estimates these operators directly from data, and uses their spectral properties for clustering and reconstruction. The reconstruction of transition densities (and the graphon under reversibility) follows from the modeling assumption that the signal arises from such a process, combined with standard operator-theoretic results; no step reduces by construction to a fitted parameter, self-defined quantity, or load-bearing self-citation chain. The central claims remain conditional on the stated modeling link and are not forced by internal redefinitions or renamings of known results.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests on standard definitions from graphon theory and transfer-operator literature; the main addition is the data-driven estimation procedure.

axioms (1)

standard math Graphons are the limit objects of convergent sequences of graphs as size tends to infinity.
Invoked when defining the random walk process on the graphon.

pith-pipeline@v0.9.0 · 5703 in / 1157 out tokens · 51479 ms · 2026-05-19T03:07:55.053414+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages

[1]

T. Ando. On compactness of integral operators. Indagationes Mathematicae, 24(2):235–239, 1962

work page 1962
[2]

Athreya and A

S. Athreya and A. R¨ ollin. Dense graph limits under respondent-driven sampling. The Annals of Applied Probability, 26(4):2193–2210, 2016. doi:10.1214/15-AAP1144

work page doi:10.1214/15-aap1144 2016
[3]

J.-P. Aubin. Applied functional analysis . John Wiley & Sons, New York, 2000

work page 2000
[4]

Avella-Medina, F

M. Avella-Medina, F. Parise, M. T. Schaub, and S. Segarra. Centrality measures for graphons: Accounting for uncertainty in networks. IEEE Transactions on Network Science and Engi- neering, 7(1):520–537, 2020. doi:10.1109/TNSE.2018.2884235

work page doi:10.1109/tnse.2018.2884235 2020
[5]

Banisch and P

R. Banisch and P. Koltai. Understanding the geometry of transport: Diffusion maps for Lagrangian trajectory data unravel coherent sets. Chaos: An Interdisciplinary Journal of Nonlinear Science, 27(3):035804, 2017. doi:10.1063/1.4971788

work page doi:10.1063/1.4971788 2017
[6]

C. M. Bishop. Pattern Recognition and Machine Learning. Springer, New York, 2006

work page 2006
[7]

Bittracher, P

A. Bittracher, P. Koltai, S. Klus, R. Banisch, M. Dellnitz, and C. Sch¨ utte. Transition manifolds of complex metastable systems: Theory and data-driven computation of effective dynamics. Journal of Nonlinear Science , 28(2):471–512, 2018. doi:10.1007/s00332-017-9415-0

work page doi:10.1007/s00332-017-9415-0 2018
[8]

Bonnet, N

B. Bonnet, N. P. Duteil, and M. Sigalotti. Consensus formation in first-order graphon mod- els with time-varying topologies. Mathematical Models and Methods in Applied Sciences , 32(11):2121–2188, 2022. doi:10.1142/S0218202522500518. 20

work page doi:10.1142/s0218202522500518 2022
[9]

Bramburger, M

J. Bramburger, M. Holzer, and J. Williams. Persistence of steady-states for dynamical systems on large networks. arXiv preprint arXiv:2402.09276 , 2024

work page arXiv 2024
[10]

J. J. Bramburger and G. Fantuzzi. Auxiliary functions as Koopman observables: Data-driven analysis of dynamical systems via polynomial optimization. Journal of Nonlinear Science , 34(1):8, 2024

work page 2024
[11]

E. B. Davies. Metastable states of symmetric Markov semigroups I. Proceedings of the London Mathematical Society, s3-45(1):133–150, 1982. doi:10.1112/plms/s3-45.1.133

work page doi:10.1112/plms/s3-45.1.133 1982
[12]

E. B. Davies. Metastable states of symmetric Markov semigroups II. Journal of the London Mathematical Society, s2-26(3):541–556, 1982. doi:10.1112/jlms/s2-26.3.541

work page doi:10.1112/jlms/s2-26.3.541 1982
[13]

Dellnitz and O

M. Dellnitz and O. Junge. On the approximation of complicated dynamical behavior. SIAM Journal on Numerical Analysis , 36(2):491–515, 1999

work page 1999
[14]

S. P. Eveson. Compactness criteria for integral operators in L∞ and L1 spaces. Proceedings of the American Mathematical Society , 123(12):3709–3716, 1995

work page 1995
[15]

A gaussian mixture distribution-based adaptive sampling method for physics-informed neural networks.Engineering Applications of Artificial Intelligence, 135:108770, 2024

G. Froyland. An analytic framework for identifying finite-time coherent sets in time-dependent dynamical systems. Physica D: Nonlinear Phenomena , 250:1–19, 2013. doi:10.1016/j. physd.2013.01.013

work page doi:10.1016/j 2013
[16]

Froyland, M

G. Froyland, M. Kalia, and P. Koltai. Spectral clustering of time-evolving networks using the inflated dynamic Laplacian for graphs, 2024. URL: https://arxiv.org/abs/2409.11984, arXiv:2409.11984

work page arXiv 2024
[17]

Froyland, N

G. Froyland, N. Santitissadeekorn, and A. Monahan. Transport in time-dependent dynamical systems: Finite-time coherent sets. Chaos: An Interdisciplinary Journal of Nonlinear Science , 20(4):043116, 2010. doi:10.1063/1.3502450

work page doi:10.1063/1.3502450 2010
[18]

Glasscock

D. Glasscock. What is... a graphon. Notices of the AMS , 62(1):46–48, 2015

work page 2015
[19]

I. G. Graham and I. H. Sloan. On the compactness of certain integral operators. Journal of Mathematical Analysis and Applications , 68(2):580–594, 1979

work page 1979
[20]

J. J. Stachurski. Economic Dynamics: Theory and Computation . MIT Press, 2nd edition, 2022

work page 2022
[21]

S. Janson. Graphons, cut norm and distance, couplings and rearrangements , volume 4 of New York Journal of Mathematics . State University of New York, University at Albany, Albany, NY, 2013

work page 2013
[22]

Klus and N

S. Klus and N. D. Conrad. Dynamical systems and complex networks: A Koopman operator perspective. Journal of Physics: Complexity , 2024. doi:10.1088/2632-072X/ad9e60

work page doi:10.1088/2632-072x/ad9e60 2024
[23]

S. Klus, B. E. Husic, M. Mollenhauer, and F. No´ e. Kernel methods for detecting coherent structures in dynamical data. Chaos, 2019. doi:10.1063/1.5100267

work page doi:10.1063/1.5100267 2019
[24]

S. Klus, P. Koltai, and C. Sch¨ utte. On the numerical approximation of the Perron–Frobenius and Koopman operator. Journal of Computational Dynamics , 3(1):51–79, 2016. doi:10. 3934/jcd.2016003

work page 2016
[25]

Klus and M

S. Klus and M. Trower. Transfer operators on graphs: Spectral clustering and beyond. Journal of Physics: Complexity , 5(1):015014, 2024. doi:10.1088/2632-072X/ad28fe. 21

work page doi:10.1088/2632-072x/ad28fe 2024
[26]

P. Koltai. Efficient approximation methods for the global long-term behavior of dynamical systems – Theory, algorithms and examples . PhD thesis, Technische Universit¨ at M¨ unchen, 2010

work page 2010
[27]

Koltai, H

P. Koltai, H. Wu, F. No´ e, and C. Sch¨ utte. Optimal data-driven estimation of generalized Markov state models for non-equilibrium dynamics. Computation, 6(1), 2018. doi:10.3390/ computation6010022

work page 2018
[28]

Korda and I

M. Korda and I. Mezi´ c. On convergence of extended dynamic mode decomposition to the Koopman operator. Journal of Nonlinear Science , 28(2):687–710, 2018

work page 2018
[29]

Lasota and M

A. Lasota and M. C. Mackey. Chaos, fractals, and noise: Stochastic aspects of dynamics , volume 97 of Applied Mathematical Sciences. Springer, New York, 2nd edition, 1994

work page 1994
[30]

Leli` evre, F

T. Leli` evre, F. Nier, and G. A. Pavliotis. Optimal non-reversible linear drift for the convergence to equilibrium of a diffusion. Journal of Statistical Physics , 152(2):237–274, 2013. doi:10. 1007/s10955-013-0769-x

work page 2013
[31]

R. Levie. Convergence and stability of graph convolutional networks on large random graphs. Advances in Neural Information Processing Systems , 33:21512–21523, 2020

work page 2020
[32]

R. Levie. A graphon-signal analysis of graph neural networks. Advances in Neural Information Processing Systems, 36:64482–64525, 2023

work page 2023
[33]

Lov´ asz

L. Lov´ asz. Random walks on graphs: A survey.Combinatorics, Paul Erd˝ os is Eighty, 2, 1993

work page 1993
[34]

Lov´ asz

L. Lov´ asz. Large networks and graph limits , volume 60. American Mathematical Sociecty, 2012

work page 2012
[35]

Lov´ asz and B

L. Lov´ asz and B. Szegedy. Limits of dense graph sequences.Journal of Combinatorial Theory, Series B, 96(6):933–957, 2006. doi:10.1016/j.jctb.2006.05.002

work page doi:10.1016/j.jctb.2006.05.002 2006
[36]

I. Mezi´ c. Spectral properties of dynamical systems, model reduction and decompositions. Nonlinear Dynamics, 41(1):309–325, 2005. doi:10.1007/s11071-005-2824-x

work page doi:10.1007/s11071-005-2824-x 2005
[37]

Mollenhauer, I

M. Mollenhauer, I. Schuster, S. Klus, and C. Sch¨ utte. Singular value decomposition of op- erators on reproducing kernel Hilbert spaces. In Advances in Dynamics, Optimization and Computation, pages 109–131, Cham, 2020. Springer. doi:10.1007/978-3-030-51264-4_5

work page doi:10.1007/978-3-030-51264-4_5 2020
[38]

M. W. Morency and G. Leus. Graphon filters: Graph signal processing in the limit. IEEE Transactions on Signal Processing, 69:1740–1754, 2021

work page 2021
[39]

A. M. Neuman and J. J. Bramburger. Transferability of graph neural networks using graphon and sampling theories. arXiv preprint arXiv:2307.13206 , 2023

work page arXiv 2023
[40]

A. Ng, M. Jordan, and Y. Weiss. On spectral clustering: Analysis and an algorithm. Advances in Neural Information Processing Systems , 14, 04 2002

work page 2002
[41]

No´ e and C

F. No´ e and C. Clementi. Kinetic distance and kinetic maps from molecular dynamics simulation. Journal of Chemical Theory and Computation , 11(10):5002–5011, 2015. doi: 10.1021/acs.jctc.5b00553

work page doi:10.1021/acs.jctc.5b00553 2015
[42]

No´ e and F

F. No´ e and F. N¨ uske. A variational approach to modeling slow processes in stochastic dynam- ical systems. Multiscale Modeling & Simulation , 11(2):635–655, 2013. 22

work page 2013
[43]

Petit, R

J. Petit, R. Lambiotte, and T. Carletti. Random walks on dense graphs and graphons. SIAM Journal on Applied Mathematics , 81(6):2323–2345, 2021. doi:10.1137/20M1339246

work page doi:10.1137/20m1339246 2021
[44]

L. Ruiz, L. Chamon, and A. Ribeiro. Graphon neural networks and the transferability of graph neural networks. Advances in Neural Information Processing Systems , 33:1702–1712, 2020

work page 2020
[45]

L. Ruiz, L. Chamon, and A. Ribeiro. Transferability properties of graph neural networks. IEEE Transactions on Signal Processing , 71:3474–3489, 2023

work page 2023
[46]

L. Ruiz, L. F. O. Chamon, and A. Ribeiro. Graphon signal processing. IEEE Transactions on Signal Processing, 69:4961–4976, 2021. doi:10.1109/TSP.2021.3106857

work page doi:10.1109/tsp.2021.3106857 2021
[47]

L. Ruiz, Z. Wang, and A. Ribeiro. Graphon and graph neural network stability. In 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 5255–5259. IEEE, 2021

work page 2021
[48]

Sch¨ utte and M

C. Sch¨ utte and M. Sarich. Metastability and Markov State Models in Molecular Dynamics: Modeling, Analysis, Algorithmic Approaches. Number 24 in Courant Lecture Notes. American Mathematical Society, 2013

work page 2013
[49]

Trower, N

M. Trower, N. D. Conrad, and S. Klus. Clustering time-evolving networks using the spatiotem- poral graph Laplacian. Chaos, 35(1):013126, 2025. doi:10.1063/5.0228419

work page doi:10.1063/5.0228419 2025
[50]

S. M. Ulam. A Collection of Mathematical Problems . Interscience Publisher NY, 1960

work page 1960
[51]

von Luxburg

U. von Luxburg. A tutorial on spectral clustering. Statistics and Computing , 17(4):395–416,

work page
[52]

doi:10.1007/s11222-007-9033-z

work page doi:10.1007/s11222-007-9033-z
[53]

D. Werner. Funktionalanalysis. Springer, Berlin, Heidelberg, 2011

work page 2011
[54]

M. O. Williams, I. G. Kevrekidis, and C. W. Rowley. A data-driven approximation of the Koopman operator: Extending dynamic mode decomposition. Journal of Nonlinear Science , 25(6):1307–1346, 2015. doi:10.1007/s00332-015-9258-5. A Spectral decompositions of compact operators For the sake of completeness, we will briefly review spectral properties of compac...

work page doi:10.1007/s00332-015-9258-5 2015

[1] [1]

T. Ando. On compactness of integral operators. Indagationes Mathematicae, 24(2):235–239, 1962

work page 1962

[2] [2]

Athreya and A

S. Athreya and A. R¨ ollin. Dense graph limits under respondent-driven sampling. The Annals of Applied Probability, 26(4):2193–2210, 2016. doi:10.1214/15-AAP1144

work page doi:10.1214/15-aap1144 2016

[3] [3]

J.-P. Aubin. Applied functional analysis . John Wiley & Sons, New York, 2000

work page 2000

[4] [4]

Avella-Medina, F

M. Avella-Medina, F. Parise, M. T. Schaub, and S. Segarra. Centrality measures for graphons: Accounting for uncertainty in networks. IEEE Transactions on Network Science and Engi- neering, 7(1):520–537, 2020. doi:10.1109/TNSE.2018.2884235

work page doi:10.1109/tnse.2018.2884235 2020

[5] [5]

Banisch and P

R. Banisch and P. Koltai. Understanding the geometry of transport: Diffusion maps for Lagrangian trajectory data unravel coherent sets. Chaos: An Interdisciplinary Journal of Nonlinear Science, 27(3):035804, 2017. doi:10.1063/1.4971788

work page doi:10.1063/1.4971788 2017

[6] [6]

C. M. Bishop. Pattern Recognition and Machine Learning. Springer, New York, 2006

work page 2006

[7] [7]

Bittracher, P

A. Bittracher, P. Koltai, S. Klus, R. Banisch, M. Dellnitz, and C. Sch¨ utte. Transition manifolds of complex metastable systems: Theory and data-driven computation of effective dynamics. Journal of Nonlinear Science , 28(2):471–512, 2018. doi:10.1007/s00332-017-9415-0

work page doi:10.1007/s00332-017-9415-0 2018

[8] [8]

Bonnet, N

B. Bonnet, N. P. Duteil, and M. Sigalotti. Consensus formation in first-order graphon mod- els with time-varying topologies. Mathematical Models and Methods in Applied Sciences , 32(11):2121–2188, 2022. doi:10.1142/S0218202522500518. 20

work page doi:10.1142/s0218202522500518 2022

[9] [9]

Bramburger, M

J. Bramburger, M. Holzer, and J. Williams. Persistence of steady-states for dynamical systems on large networks. arXiv preprint arXiv:2402.09276 , 2024

work page arXiv 2024

[10] [10]

J. J. Bramburger and G. Fantuzzi. Auxiliary functions as Koopman observables: Data-driven analysis of dynamical systems via polynomial optimization. Journal of Nonlinear Science , 34(1):8, 2024

work page 2024

[11] [11]

E. B. Davies. Metastable states of symmetric Markov semigroups I. Proceedings of the London Mathematical Society, s3-45(1):133–150, 1982. doi:10.1112/plms/s3-45.1.133

work page doi:10.1112/plms/s3-45.1.133 1982

[12] [12]

E. B. Davies. Metastable states of symmetric Markov semigroups II. Journal of the London Mathematical Society, s2-26(3):541–556, 1982. doi:10.1112/jlms/s2-26.3.541

work page doi:10.1112/jlms/s2-26.3.541 1982

[13] [13]

Dellnitz and O

M. Dellnitz and O. Junge. On the approximation of complicated dynamical behavior. SIAM Journal on Numerical Analysis , 36(2):491–515, 1999

work page 1999

[14] [14]

S. P. Eveson. Compactness criteria for integral operators in L∞ and L1 spaces. Proceedings of the American Mathematical Society , 123(12):3709–3716, 1995

work page 1995

[15] [15]

A gaussian mixture distribution-based adaptive sampling method for physics-informed neural networks.Engineering Applications of Artificial Intelligence, 135:108770, 2024

G. Froyland. An analytic framework for identifying finite-time coherent sets in time-dependent dynamical systems. Physica D: Nonlinear Phenomena , 250:1–19, 2013. doi:10.1016/j. physd.2013.01.013

work page doi:10.1016/j 2013

[16] [16]

Froyland, M

G. Froyland, M. Kalia, and P. Koltai. Spectral clustering of time-evolving networks using the inflated dynamic Laplacian for graphs, 2024. URL: https://arxiv.org/abs/2409.11984, arXiv:2409.11984

work page arXiv 2024

[17] [17]

Froyland, N

G. Froyland, N. Santitissadeekorn, and A. Monahan. Transport in time-dependent dynamical systems: Finite-time coherent sets. Chaos: An Interdisciplinary Journal of Nonlinear Science , 20(4):043116, 2010. doi:10.1063/1.3502450

work page doi:10.1063/1.3502450 2010

[18] [18]

Glasscock

D. Glasscock. What is... a graphon. Notices of the AMS , 62(1):46–48, 2015

work page 2015

[19] [19]

I. G. Graham and I. H. Sloan. On the compactness of certain integral operators. Journal of Mathematical Analysis and Applications , 68(2):580–594, 1979

work page 1979

[20] [20]

J. J. Stachurski. Economic Dynamics: Theory and Computation . MIT Press, 2nd edition, 2022

work page 2022

[21] [21]

S. Janson. Graphons, cut norm and distance, couplings and rearrangements , volume 4 of New York Journal of Mathematics . State University of New York, University at Albany, Albany, NY, 2013

work page 2013

[22] [22]

Klus and N

S. Klus and N. D. Conrad. Dynamical systems and complex networks: A Koopman operator perspective. Journal of Physics: Complexity , 2024. doi:10.1088/2632-072X/ad9e60

work page doi:10.1088/2632-072x/ad9e60 2024

[23] [23]

S. Klus, B. E. Husic, M. Mollenhauer, and F. No´ e. Kernel methods for detecting coherent structures in dynamical data. Chaos, 2019. doi:10.1063/1.5100267

work page doi:10.1063/1.5100267 2019

[24] [24]

S. Klus, P. Koltai, and C. Sch¨ utte. On the numerical approximation of the Perron–Frobenius and Koopman operator. Journal of Computational Dynamics , 3(1):51–79, 2016. doi:10. 3934/jcd.2016003

work page 2016

[25] [25]

Klus and M

S. Klus and M. Trower. Transfer operators on graphs: Spectral clustering and beyond. Journal of Physics: Complexity , 5(1):015014, 2024. doi:10.1088/2632-072X/ad28fe. 21

work page doi:10.1088/2632-072x/ad28fe 2024

[26] [26]

P. Koltai. Efficient approximation methods for the global long-term behavior of dynamical systems – Theory, algorithms and examples . PhD thesis, Technische Universit¨ at M¨ unchen, 2010

work page 2010

[27] [27]

Koltai, H

P. Koltai, H. Wu, F. No´ e, and C. Sch¨ utte. Optimal data-driven estimation of generalized Markov state models for non-equilibrium dynamics. Computation, 6(1), 2018. doi:10.3390/ computation6010022

work page 2018

[28] [28]

Korda and I

M. Korda and I. Mezi´ c. On convergence of extended dynamic mode decomposition to the Koopman operator. Journal of Nonlinear Science , 28(2):687–710, 2018

work page 2018

[29] [29]

Lasota and M

A. Lasota and M. C. Mackey. Chaos, fractals, and noise: Stochastic aspects of dynamics , volume 97 of Applied Mathematical Sciences. Springer, New York, 2nd edition, 1994

work page 1994

[30] [30]

Leli` evre, F

T. Leli` evre, F. Nier, and G. A. Pavliotis. Optimal non-reversible linear drift for the convergence to equilibrium of a diffusion. Journal of Statistical Physics , 152(2):237–274, 2013. doi:10. 1007/s10955-013-0769-x

work page 2013

[31] [31]

R. Levie. Convergence and stability of graph convolutional networks on large random graphs. Advances in Neural Information Processing Systems , 33:21512–21523, 2020

work page 2020

[32] [32]

R. Levie. A graphon-signal analysis of graph neural networks. Advances in Neural Information Processing Systems, 36:64482–64525, 2023

work page 2023

[33] [33]

Lov´ asz

L. Lov´ asz. Random walks on graphs: A survey.Combinatorics, Paul Erd˝ os is Eighty, 2, 1993

work page 1993

[34] [34]

Lov´ asz

L. Lov´ asz. Large networks and graph limits , volume 60. American Mathematical Sociecty, 2012

work page 2012

[35] [35]

Lov´ asz and B

L. Lov´ asz and B. Szegedy. Limits of dense graph sequences.Journal of Combinatorial Theory, Series B, 96(6):933–957, 2006. doi:10.1016/j.jctb.2006.05.002

work page doi:10.1016/j.jctb.2006.05.002 2006

[36] [36]

I. Mezi´ c. Spectral properties of dynamical systems, model reduction and decompositions. Nonlinear Dynamics, 41(1):309–325, 2005. doi:10.1007/s11071-005-2824-x

work page doi:10.1007/s11071-005-2824-x 2005

[37] [37]

Mollenhauer, I

M. Mollenhauer, I. Schuster, S. Klus, and C. Sch¨ utte. Singular value decomposition of op- erators on reproducing kernel Hilbert spaces. In Advances in Dynamics, Optimization and Computation, pages 109–131, Cham, 2020. Springer. doi:10.1007/978-3-030-51264-4_5

work page doi:10.1007/978-3-030-51264-4_5 2020

[38] [38]

M. W. Morency and G. Leus. Graphon filters: Graph signal processing in the limit. IEEE Transactions on Signal Processing, 69:1740–1754, 2021

work page 2021

[39] [39]

A. M. Neuman and J. J. Bramburger. Transferability of graph neural networks using graphon and sampling theories. arXiv preprint arXiv:2307.13206 , 2023

work page arXiv 2023

[40] [40]

A. Ng, M. Jordan, and Y. Weiss. On spectral clustering: Analysis and an algorithm. Advances in Neural Information Processing Systems , 14, 04 2002

work page 2002

[41] [41]

No´ e and C

F. No´ e and C. Clementi. Kinetic distance and kinetic maps from molecular dynamics simulation. Journal of Chemical Theory and Computation , 11(10):5002–5011, 2015. doi: 10.1021/acs.jctc.5b00553

work page doi:10.1021/acs.jctc.5b00553 2015

[42] [42]

No´ e and F

F. No´ e and F. N¨ uske. A variational approach to modeling slow processes in stochastic dynam- ical systems. Multiscale Modeling & Simulation , 11(2):635–655, 2013. 22

work page 2013

[43] [43]

Petit, R

J. Petit, R. Lambiotte, and T. Carletti. Random walks on dense graphs and graphons. SIAM Journal on Applied Mathematics , 81(6):2323–2345, 2021. doi:10.1137/20M1339246

work page doi:10.1137/20m1339246 2021

[44] [44]

L. Ruiz, L. Chamon, and A. Ribeiro. Graphon neural networks and the transferability of graph neural networks. Advances in Neural Information Processing Systems , 33:1702–1712, 2020

work page 2020

[45] [45]

L. Ruiz, L. Chamon, and A. Ribeiro. Transferability properties of graph neural networks. IEEE Transactions on Signal Processing , 71:3474–3489, 2023

work page 2023

[46] [46]

L. Ruiz, L. F. O. Chamon, and A. Ribeiro. Graphon signal processing. IEEE Transactions on Signal Processing, 69:4961–4976, 2021. doi:10.1109/TSP.2021.3106857

work page doi:10.1109/tsp.2021.3106857 2021

[47] [47]

L. Ruiz, Z. Wang, and A. Ribeiro. Graphon and graph neural network stability. In 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 5255–5259. IEEE, 2021

work page 2021

[48] [48]

Sch¨ utte and M

C. Sch¨ utte and M. Sarich. Metastability and Markov State Models in Molecular Dynamics: Modeling, Analysis, Algorithmic Approaches. Number 24 in Courant Lecture Notes. American Mathematical Society, 2013

work page 2013

[49] [49]

Trower, N

M. Trower, N. D. Conrad, and S. Klus. Clustering time-evolving networks using the spatiotem- poral graph Laplacian. Chaos, 35(1):013126, 2025. doi:10.1063/5.0228419

work page doi:10.1063/5.0228419 2025

[50] [50]

S. M. Ulam. A Collection of Mathematical Problems . Interscience Publisher NY, 1960

work page 1960

[51] [51]

von Luxburg

U. von Luxburg. A tutorial on spectral clustering. Statistics and Computing , 17(4):395–416,

work page

[52] [52]

doi:10.1007/s11222-007-9033-z

work page doi:10.1007/s11222-007-9033-z

[53] [53]

D. Werner. Funktionalanalysis. Springer, Berlin, Heidelberg, 2011

work page 2011

[54] [54]

M. O. Williams, I. G. Kevrekidis, and C. W. Rowley. A data-driven approximation of the Koopman operator: Extending dynamic mode decomposition. Journal of Nonlinear Science , 25(6):1307–1346, 2015. doi:10.1007/s00332-015-9258-5. A Spectral decompositions of compact operators For the sake of completeness, we will briefly review spectral properties of compac...

work page doi:10.1007/s00332-015-9258-5 2015