pith. sign in

arxiv: 2606.29098 · v1 · pith:GULZ2RWFnew · submitted 2026-06-27 · 📊 stat.ML · cs.LG· cs.SI· eess.SP· q-bio.NC

Connectivity Estimation using Stochastic Graph Heat Modelling

Stephan Goerttler , Min Wu , Fei He This is my paper

Pith reviewed 2026-06-30 08:02 UTC · model grok-4.3

classification 📊 stat.ML cs.LGcs.SIeess.SPq-bio.NC
keywords connectivity estimationgraph heat modellingneurophysiological dataspatial structureregularizationdirected connectivitystochastic modellingbrain networks
0
0 comments X

The pith

Relaxing noise assumptions and adding regularization to stochastic graph heat models improves estimation of directed connectivity in neurophysiological recordings.

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

The paper extends an earlier noise-driven heat modelling approach on graphs to estimate brain connectivity. It relaxes prior noise assumptions, introduces regularization for robustness, and adds a simulation procedure for controlled evaluation. The resulting estimator is explicitly model-based, dynamic, multivariate, and directed. When applied to real-world datasets in two separate experiments, it recovers meaningful spatial structures. The explicit formulation is intended to increase interpretability of graph-based methods more broadly.

Core claim

By relaxing earlier noise assumptions and incorporating regularization, the stochastic graph heat model yields a connectivity estimator that captures meaningful spatial structure in neurophysiological recordings, as demonstrated through simulation studies and application to multiple real-world datasets in two experiments.

What carries the argument

The stochastic graph heat model with relaxed noise assumptions and added regularization, which treats connectivity as the structure governing noise-driven heat diffusion on graphs.

If this is right

  • The estimator produces directed, dynamic, and multivariate connectivity values rather than undirected summaries.
  • Regularization makes the estimates more stable when applied to noisy real recordings.
  • The simulation procedure provides a controlled way to characterize performance before real-data use.
  • An explicit model formulation supports direct interpretation of the estimated connectivities.
  • The same framework can be applied across different neurophysiological recording modalities.

Where Pith is reading between the lines

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

  • The approach could be tested on non-brain spatial datasets where similar diffusion-like processes are plausible.
  • If the estimated connectivities prove stable across subjects, they might serve as features for downstream clinical classification tasks.
  • Further work could examine how the regularization parameter trades off bias and variance in the recovered graphs.

Load-bearing premise

The stochastic graph heat model, once its noise assumptions are relaxed, accurately reflects the generative process that produced the neurophysiological recordings.

What would settle it

If the connectivities estimated by the model show no alignment with independently known anatomical or functional structures in the same real-world datasets, the claim that the model captures meaningful spatial structure would be refuted.

Figures

Figures reproduced from arXiv: 2606.29098 by Fei He, Min Wu, Stephan Goerttler.

Figure 1
Figure 1. Figure 1: Illustration of the proposed stochastic graph heat model. The model is shown for a six-node undirected (symmetric) graph. Heat dif￾fuses from each node to its neighbours (red arrows); for clarity, diffusion is illustrated only for node 3. The amount of transferred heat depends on both the connection strength (line thickness) and the temperature gradient between the nodes. Noise is continuously added to the… view at source ↗
Figure 2
Figure 2. Figure 2: FMRI connectivity estimation for a single EEG sensor pair. (A) Spherical neighbourhoods in the fMRI volume, centred on the cortical projections of each EEG sensor, consist of voxels with time-varying signals. The signals within each neighbourhood are spatially averaged to yield (B) a time series for each sensor. Pairwise connectivity is then computed between sensors using Pearson correlation. C. Heat model… view at source ↗
Figure 3
Figure 3. Figure 3: Insertion of spurious correlation between two arbitrary nodes i and k. The direct connection between i and k is removed, while the connection between i and k through an intermediate node j is increased. All other connections remain the same. configurations by generating five logarithmically spaced values for α ∈ [0.01, 0.1] and σ int ∈ [0.1, 10], while keeping σ ext = 1 fixed. Furthermore, for the artifici… view at source ↗
Figure 4
Figure 4. Figure 4: ROIs (in red) of EEG sensors for (A) Dataset III and (B) Dataset IV. Slice values were chosen to maximise the number of sensors displayed. Note that not all sensors are visible in the two-dimensional slices due to the 3D slicing geometry, and some may appear more than once. are sufficiently large. For small values of σint, the correla￾tion measure outperforms the Heat-2R measure, although the performance o… view at source ↗
Figure 5
Figure 5. Figure 5: Mean ground truth alignment of the heat-based simulation for four selected connectivity measures, across all parameter configurations for undirected (first row) and directed (second row) graphs. Red stars indicate the best performance across the four connectivity measures for each graph type. The Heat-2R measure clearly outperforms all other measures, particularly at large values of σint. measure also perf… view at source ↗
Figure 6
Figure 6. Figure 6: Scaled retrieved connectivity values over all connectivity measures for the spurious connection and the two real intermediate connections. The heat-based connectivity measures Heat-EV (blue), Heat-2 (light blue), and Heat-2R (dark blue) are closest to the ground truth connectivity values of 0 for the spurious and 0.8 for the intermediate connections. The best-performing baselines are Granger causality and … view at source ↗
Figure 7
Figure 7. Figure 7: Mean ground truth alignment of the correlation-based simu￾lation across maximum frequency values between 1 and 25. Shaded areas represent the standard error. The ground truth alignment for the correlation measure is close to 1, as intended by design. The heat connectivity measures are sensitive to the correlation structure when higher frequencies are excluded, but this sensitivity diminishes once sufficien… view at source ↗
Figure 8
Figure 8. Figure 8: presents the results of Experiment I, showing the SVM classification accuracy for the three heat-based connec￾tivity measures and nine baseline connectivity measures for Datasets I and II. The two heat-based connectivity measures Heat-EV and Heat-2R outperform all other measures on both datasets. While Heat-2R performs best on Dataset I, Heat-EV achieves the highest accuracy on Dataset II [PITH_FULL_IMAGE… view at source ↗
Figure 9
Figure 9. Figure 9: shows the results of Experiment II as violin plots of the alignment distribution across all participants for Datasets III and IV. Heat-2R connectivity exhibits the highest fMRI alignment overall, surpassed only by correlation and geometric connectivity on Dataset IV. Additionally, its align￾ment distribution is mostly positive. On Dataset III, coherence connectivity follows Heat-2R in alignment. Among the … view at source ↗
Figure 10
Figure 10. Figure 10: Parameter sensitivity analysis for connectivity-derived SVM classification accuracy for Datasets I (A) and II (B). Performance is shown as bootstrapped mean with standard error across various target frequencies. Overall, performance is higher for intermediate sampling rate values. However, the performance is sensitive to the sampling rate, particularly for Dataset I. 0.1 0.0 0.1 0.2 0.3 fMRI alignment 20 … view at source ↗
read the original abstract

A growing number of techniques leverage the spatial structures that underlie many real-world datasets. Despite these advances, the complementary task of estimating spatial structures and understanding their role within these techniques has often been overlooked. In neurophysiological data analysis specifically, numerous methods exist to estimate brain connectivity, but most are not explicitly model-based, dynamic, multivariate, or directed. To address these limitations, we previously introduced noise-driven heat modelling on graphs for neurophysiological connectivity estimation. In this study, we extend this framework by relaxing earlier noise assumptions and adding regularisation to improve robustness. We also develop a simulation procedure to characterise and evaluate our technique in a controlled setting. Finally, we demonstrate that the technique is able to capture meaningful spatial structure across two experiments, each using two real-world datasets. The explicit model formulation of our connectivity estimator has the potential to improve the interpretability of graph-based techniques across a wide range of applications. The code implementing our method is available at https://github.com/sgoerttler/Heat_Connectivity.

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 manuscript extends the authors' prior noise-driven heat modelling on graphs for neurophysiological connectivity estimation. The extension relaxes earlier noise assumptions and introduces regularization for improved robustness. A simulation procedure is developed to evaluate the method in controlled settings, and the technique is demonstrated to capture meaningful spatial structure across two experiments, each using two real-world datasets. The explicit model-based formulation is positioned as improving interpretability of graph-based techniques, with code released at the provided GitHub repository.

Significance. If the central claims hold, the work supplies a model-based, dynamic, multivariate, and directed connectivity estimator that could aid interpretability across applications. Strengths include the simulation-based evaluation framework and public code release, which support reproducibility.

major comments (2)
  1. [Real-data experiments] §Real-data experiments: The headline demonstration that the estimator captures meaningful spatial structure in the four real-world datasets rests on the assumption that the relaxed stochastic graph heat model is a sufficiently accurate generative model for the recordings. No ground-truth validation or independent check (e.g., against known anatomical pathways or alternative modalities) is provided, so it remains possible that recovered structures reflect model assumptions rather than true connectivity.
  2. [Simulation procedure] Simulation procedure section: The procedure is introduced to characterise the technique, yet the manuscript does not detail how the relaxation of noise assumptions is instantiated in the simulated data-generating process or how estimation error is quantified under the new regularization; without these specifics the controlled evaluation cannot fully substantiate the robustness claims.
minor comments (2)
  1. The abstract states that two experiments each use two real-world datasets; naming the specific datasets and the precise experimental conditions in the abstract or introduction would improve immediate clarity.
  2. The regularization parameter is listed among free parameters; its selection procedure and sensitivity analysis should be stated explicitly in the methods.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and outline the corresponding revisions.

read point-by-point responses

  1. Referee: [Real-data experiments] §Real-data experiments: The headline demonstration that the estimator captures meaningful spatial structure in the four real-world datasets rests on the assumption that the relaxed stochastic graph heat model is a sufficiently accurate generative model for the recordings. No ground-truth validation or independent check (e.g., against known anatomical pathways or alternative modalities) is provided, so it remains possible that recovered structures reflect model assumptions rather than true connectivity.

    Authors: We agree that the real-data results are conditional on the model being a reasonable approximation. The experiments are designed to illustrate recovery of spatially coherent patterns that align with domain knowledge in the chosen datasets, rather than to claim recovery of ground-truth connectivity. In the revision we will add an explicit limitations paragraph stating this assumption and the lack of independent validation, thereby clarifying the scope of the claims. revision: partial

  2. Referee: [Simulation procedure] Simulation procedure section: The procedure is introduced to characterise the technique, yet the manuscript does not detail how the relaxation of noise assumptions is instantiated in the simulated data-generating process or how estimation error is quantified under the new regularization; without these specifics the controlled evaluation cannot fully substantiate the robustness claims.

    Authors: The observation is correct; the current text omits these implementation details. We will expand the simulation section to specify (i) the exact form of the relaxed noise process used to generate the synthetic data and (ii) the quantitative error measures applied when evaluating the regularized estimator. These additions will allow readers to reproduce and assess the robustness evaluation. revision: yes

Circularity Check

0 steps flagged

Minor self-citation to prior model; derivation and empirical claims remain independent

full rationale

The paper cites its authors' prior introduction of the base noise-driven heat model but presents an explicit extension (relaxed noise assumptions, added regularization, new simulation procedure, and fresh real-data experiments). No equations or steps are shown that reduce new predictions to fitted inputs by construction, import uniqueness via self-citation, or smuggle ansatzes. The self-citation supplies background context rather than load-bearing justification for the central claims about capturing spatial structure. The derivation chain for the extended estimator and its validation is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Based on abstract only; the central modelling approach rests on domain assumptions about graph heat diffusion representing brain dynamics and the benefit of added regularization.

free parameters (1)
  • regularization parameter
    Introduced to improve robustness; value likely chosen or fitted but not specified in abstract.
axioms (1)
  • domain assumption The graph heat model represents the underlying dynamics of neurophysiological signals after relaxing earlier noise assumptions.
    Core premise of the connectivity estimator.

pith-pipeline@v0.9.1-grok · 5714 in / 1101 out tokens · 32828 ms · 2026-06-30T08:02:13.902220+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

35 extracted references

  1. [1]

    A comprehensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, “A comprehensive survey on graph neural networks,”IEEE transactions on neural networks and learning systems, vol. 32, no. 1, pp. 4–24, 2020

    2020

  2. [2]

    Learning heat diffu- sion graphs,

    D. Thanou, X. Dong, D. Kressner, and P. Frossard, “Learning heat diffu- sion graphs,”IEEE Transactions on Signal and Information Processing over Networks, vol. 3, no. 3, pp. 484–499, 2017

    2017

  3. [3]

    Understanding concepts in graph signal processing for neurophysiological signal analysis,

    S. Goerttler, M. Wu, and F. He, “Understanding concepts in graph signal processing for neurophysiological signal analysis,” inMachine Learning Applications in Medicine and Biology. Springer, 2024, pp. 1–41

    2024

  4. [4]

    Graph neural network-based eeg classification: A survey,

    D. Klepl, M. Wu, and F. He, “Graph neural network-based eeg classification: A survey,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 32, pp. 493–503, 2024

    2024

  5. [5]

    Phase lag index: assess- ment of functional connectivity from multi channel eeg and meg with diminished bias from common sources,

    C. J. Stam, G. Nolte, and A. Daffertshofer, “Phase lag index: assess- ment of functional connectivity from multi channel eeg and meg with diminished bias from common sources,”Human brain mapping, vol. 28, no. 11, pp. 1178–1193, 2007

    2007

  6. [6]

    Measuring phase synchrony in brain signals,

    J.-P. Lachaux, E. Rodriguez, J. Martinerie, and F. J. Varela, “Measuring phase synchrony in brain signals,”Human brain mapping, vol. 8, no. 4, pp. 194–208, 1999

    1999

  7. [7]

    Investigating the electrophysiological basis of resting state networks using magnetoen- cephalography,

    M. J. Brookes, M. Woolrich, H. Luckhoo, D. Price, J. R. Hale, M. C. Stephenson, G. R. Barnes, S. M. Smith, and P. G. Morris, “Investigating the electrophysiological basis of resting state networks using magnetoen- cephalography,”Proceedings of the National Academy of Sciences, vol. 108, no. 40, pp. 16 783–16 788, 2011

    2011

  8. [8]

    A tutorial review of functional con- nectivity analysis methods and their interpretational pitfalls,

    A. M. Bastos and J.-M. Schoffelen, “A tutorial review of functional con- nectivity analysis methods and their interpretational pitfalls,”Frontiers in systems neuroscience, vol. 9, p. 175, 2016

    2016

  9. [9]

    Directed brain connectivity identifies widespread functional network abnormalities in parkinson’s disease,

    M. Mijalkov, G. V olpe, and J. B. Pereira, “Directed brain connectivity identifies widespread functional network abnormalities in parkinson’s disease,”Cerebral cortex, vol. 32, no. 3, pp. 593–607, 2022

    2022

  10. [10]

    Stochastic graph heat modelling for diffusion-based connectivity retrieval,

    S. Goerttler, F. He, and M. Wu, “Stochastic graph heat modelling for diffusion-based connectivity retrieval,” in2024 46th Annual Interna- 10 STEPHAN GOERTTLERet al.: CONNECTIVITY ESTIMATION USING STOCHASTIC GRAPH HEAT MODELLING TABLE II EXPERIMENTII: FMRIALIGNMENT OF CONNECTIVITY MEASURES. ALSO DISPLAYED ARE THE COMPUTATION TIME PER SAMPLE AND THEp-VALUE ...

    2024

  11. [11]

    Human brain networks func- tion in connectome-specific harmonic waves,

    S. Atasoy, I. Donnelly, and J. Pearson, “Human brain networks func- tion in connectome-specific harmonic waves,”Nature communications, vol. 7, no. 1, p. 10340, 2016

    2016

  12. [12]

    Functional importance of noise in neuronal information processing,

    D. Guo, M. Perc, T. Liu, and D. Yao, “Functional importance of noise in neuronal information processing,”Europhysics Letters, vol. 124, no. 5, p. 50001, 2018

    2018

  13. [13]

    Prediction of ni ˜no 3 sea surface temper- atures using linear inverse modeling,

    C. Penland and T. Magorian, “Prediction of ni ˜no 3 sea surface temper- atures using linear inverse modeling,”Journal of Climate, vol. 6, no. 6, pp. 1067–1076, 1993

    1993

  14. [14]

    Dynamic causal modelling,

    K. J. Friston, L. Harrison, and W. Penny, “Dynamic causal modelling,” Neuroimage, vol. 19, no. 4, pp. 1273–1302, 2003

    2003

  15. [15]

    Dynamic causal mod- elling: a critical review of the biophysical and statistical foundations,

    J. Daunizeau, O. David, and K. E. Stephan, “Dynamic causal mod- elling: a critical review of the biophysical and statistical foundations,” Neuroimage, vol. 58, no. 2, pp. 312–322, 2011

    2011

  16. [16]

    Emerging concepts for the dynamical organization of resting-state activity in the brain,

    G. Deco, V . K. Jirsa, and A. R. McIntosh, “Emerging concepts for the dynamical organization of resting-state activity in the brain,”Nature reviews neuroscience, vol. 12, no. 1, pp. 43–56, 2011

    2011

  17. [17]

    Dynamic models of large-scale brain activity,

    M. Breakspear, “Dynamic models of large-scale brain activity,”Nature neuroscience, vol. 20, no. 3, pp. 340–352, 2017

    2017

  18. [18]

    Dyconnmap: Dy- namic connectome mapping—a neuroimaging python module,

    A. D. Marimpis, S. I. Dimitriadis, and R. Goebel, “Dyconnmap: Dy- namic connectome mapping—a neuroimaging python module,”Human Brain Mapping, vol. 42, no. 15, pp. 4909–4939, 2021

    2021

  19. [19]

    An improved index of phase-synchronization for electrophys- iological data in the presence of volume-conduction, noise and sample- size bias,

    M. Vinck, R. Oostenveld, M. Van Wingerden, F. Battaglia, and C. M. Pennartz, “An improved index of phase-synchronization for electrophys- iological data in the presence of volume-conduction, noise and sample- size bias,”Neuroimage, vol. 55, no. 4, pp. 1548–1565, 2011

    2011

  20. [20]

    Identifying true brain interaction from eeg data using the imaginary part of coherency,

    G. Nolte, O. Bai, L. Wheaton, Z. Mari, S. V orbach, and M. Hallett, “Identifying true brain interaction from eeg data using the imaginary part of coherency,”Clinical neurophysiology, vol. 115, no. 10, pp. 2292– 2307, 2004

    2004

  21. [21]

    Estimating mutual information,

    A. Kraskov, H. St ¨ogbauer, and P. Grassberger, “Estimating mutual information,”Physical Review E—Statistical, Nonlinear, and Soft Matter Physics, vol. 69, no. 6, p. 066138, 2004

    2004

  22. [22]

    Non-parametric entropy estimation toolbox (npeet),

    G. Ver Steeg, “Non-parametric entropy estimation toolbox (npeet),” Non-parametric entropy estimation toolbox (NPEET), 2000. [Online]. Available: https://github.com/gregversteeg/NPEET

    2000

  23. [23]

    Statsmodels: econometric and statistical modeling with python

    S. Seabold and J. Perktold, “Statsmodels: econometric and statistical modeling with python.”SciPy, vol. 7, no. 1, pp. 92–96, 2010

    2010

  24. [24]

    Scikit-learn: Machine learning in python,

    F. Pedregosa, G. Varoquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V . Dubourget al., “Scikit-learn: Machine learning in python,”the Journal of machine Learning research, vol. 12, pp. 2825–2830, 2011

    2011

  25. [25]

    Python 3.11.11,

    P. S. Foundation, “Python 3.11.11,” 2024. [Online]. Available: https://www.python.org/downloads/release/python-31111/

    2024

  26. [26]

    Evoked mid-frontal activity predicts cognitive dysfunction in parkinson’s disease,

    A. Singh, R. C. Cole, A. I. Espinoza, J. R. Wessel, J. F. Cavanagh, and N. S. Narayanan, “Evoked mid-frontal activity predicts cognitive dysfunction in parkinson’s disease,”Journal of Neurology, Neurosurgery & Psychiatry, vol. 94, no. 11, pp. 945–953, 2023

    2023

  27. [27]

    A dataset of eeg recordings from: Alzheimer’s disease, frontotemporal dementia and healthy subjects,

    A. Miltiadous, K. D. Tzimourta, T. Afrantou, P. Ioannidis, N. Grigori- adis, D. G. Tsalikakis, P. Angelidis, M. G. Tsipouras, E. Glavas, N. Gi- annakeaset al., “A dataset of eeg recordings from: Alzheimer’s disease, frontotemporal dementia and healthy subjects,”OpenNeuro.[Dataset], 2023

    2023

  28. [28]

    Simultaneous and independent electroencephalog- raphy and magnetic resonance imaging: A multimodal neuroimaging dataset,

    J. Gallego-Rudolf, M. Corsi-Cabrera, L. Concha, J. Ricardo-Garcell, and E. Pasaye-Alcaraz, “Simultaneous and independent electroencephalog- raphy and magnetic resonance imaging: A multimodal neuroimaging dataset,”Data in Brief, vol. 51, p. 109661, 2023

    2023

  29. [29]

    An open-access dataset of naturalistic viewing using simultaneous eeg-fmri,

    Q. K. Telesford, E. Gonzalez-Moreira, T. Xu, Y . Tian, S. J. Colcombe, J. Cloud, B. E. Russ, A. Falchier, M. Nentwich, J. Madsenet al., “An open-access dataset of naturalistic viewing using simultaneous eeg-fmri,” Scientific Data, vol. 10, no. 1, p. 554, 2023

    2023

  30. [30]

    Improved op- timization for the robust and accurate linear registration and motion correction of brain images,

    M. Jenkinson, P. Bannister, M. Brady, and S. Smith, “Improved op- timization for the robust and accurate linear registration and motion correction of brain images,”Neuroimage, vol. 17, no. 2, pp. 825–841, 2002

    2002

  31. [31]

    Jenkinson, C

    M. Jenkinson, C. F. Beckmann, T. E. Behrens, M. W. Woolrich, and S. M. Smith, “Fsl,”Neuroimage, vol. 62, no. 2, pp. 782–790, 2012

    2012

  32. [32]

    Unbiased average age-appropriate atlases for pediatric studies,

    V . Fonov, A. C. Evans, K. Botteron, C. R. Almli, R. C. McKinstry, D. L. Collins, B. D. C. Groupet al., “Unbiased average age-appropriate atlases for pediatric studies,”Neuroimage, vol. 54, no. 1, pp. 313–327, 2011

    2011

  33. [33]

    A reproducible evaluation of ants similarity metric performance in brain image registration,

    B. B. Avants, N. J. Tustison, G. Song, P. A. Cook, A. Klein, and J. C. Gee, “A reproducible evaluation of ants similarity metric performance in brain image registration,”Neuroimage, vol. 54, no. 3, pp. 2033–2044, 2011

    2033

  34. [34]

    ANTsPy: Advanced Normalization Tools in Python,

    ANTsX Developers, “ANTsPy: Advanced Normalization Tools in Python,” https://github.com/ANTsX/ANTsPy, 2024, accessed: 2025-04- 01

    2024

  35. [35]

    Meg and eeg data analysis with mne-python,

    A. Gramfort, M. Luessi, E. Larson, D. A. Engemann, D. Strohmeier, C. Brodbeck, R. Goj, M. Jas, T. Brooks, L. Parkkonenet al., “Meg and eeg data analysis with mne-python,”Frontiers in Neuroinformatics, vol. 7, p. 267, 2013. 11 Supplemental Material:Connectivity Estimation using Stochastic Graph Heat Modelling VIII. STOCHASTICGRAPHHEATEQUATION The heat equa...

    2013