pith. sign in

arxiv: 2601.21170 · v3 · pith:7NFTIC3Xnew · submitted 2026-01-29 · 💻 cs.LG · stat.ML

The Powers of Precision: Structure-Informed Detection in Complex Systems -- From Customer Churn to Seizure Onset

Augusto Santos , Teresa Santos , Catarina Rodrigues , Jos\'e M. F. Moura This is my paper

Pith reviewed 2026-05-21 14:26 UTC · model grok-4.3

classification 💻 cs.LG stat.ML
keywords covariance matrix powersprecision matrixstructure detectionseizure onsetcustomer churncomplex systemslatent causal structuremachine learning
0
0 comments X

The pith

Powers of the empirical covariance or precision matrix capture the latent structure driving critical events like seizures and customer churn.

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

The paper develops a machine learning method to detect the onset of emergent phenomena in complex systems where the data-generating process is unknown and only partially observed. It learns an optimal feature representation by choosing the best power from the one-parameter family of estimators formed by raising the empirical covariance or precision matrix to that power. This choice tunes the representation to the hidden causal interactions that precede events such as epileptic seizures or sudden customer churn. A supervised classifier then uses the tuned representation for prediction, while the selected power also supports explainability by surfacing structural signatures. The work proves structural consistency of the estimator family and reports competitive empirical results on both seizure detection and churn prediction tasks.

Core claim

The method learns an optimal feature representation from a one-parameter family of estimators—powers of the empirical covariance or precision matrix—offering a principled way to tune in to the underlying structure driving the emergence of critical events. Structural consistency of the family is proven. Empirical results attain competitive performance on seizure detection and churn prediction. The optimal covariance power exhibits evidence of good identifiability while capturing structural signatures, reconciling predictive performance with interpretable statistical structure.

What carries the argument

the one-parameter family of powers of the empirical covariance or precision matrix, which acts as tunable estimators to extract structure-informed features from partially observed data

If this is right

  • The method attains competitive predictive performance on real datasets for seizure onset detection and customer churn prediction.
  • The optimal power parameter captures structural signatures relevant to the critical events.
  • Structural consistency holds for the full family of covariance and precision matrix powers.
  • The selected power reconciles accurate classification with interpretable statistical structure.

Where Pith is reading between the lines

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

  • The same power-selection procedure could be applied to other domains exhibiting emergent events, such as financial market crashes or epidemic spread, to test whether the family remains informative outside the two demonstrated cases.
  • Direct comparison against ground-truth network models on synthetic data with engineered causal links would provide a concrete check on whether the recovered optimal power matches the true interaction strengths.
  • The framework implies that the power parameter functions as a continuous dial for emphasizing different scales of interactions, which might be used to design new diagnostic statistics for network stability.

Load-bearing premise

The latent causal structure of the unknown and partially observed data-generating process can be effectively unveiled and harnessed by powers of the empirical covariance or precision matrix.

What would settle it

In a controlled simulation of a complex system whose true interaction structure and critical-transition mechanism are known in advance, the method selecting an optimal power that fails to recover or align with those known interactions before the transition occurs would falsify the central claim.

Figures

Figures reproduced from arXiv: 2601.21170 by Augusto Santos, Catarina Rodrigues, Jos\'e M. F. Moura, Teresa Santos.

Figure 1
Figure 1. Figure 1: Proposed approach. Top: theoretical regimes in which specific covariance powers are structurally consistent. Bottom: pipeline: data 7→ covariance 7→ power transform C β 7→ super￾vised classifier; β ⋆ is selected on a train–validation split. 5.2 Data preparation EEG seizure onset (CHB-MIT). We use the CHB-MIT scalp EEG dataset (23 bipolar chan￾nels, 256 Hz), containing 24 pediatric cases with 19 seizures. T… view at source ↗
Figure 2
Figure 2. Figure 2: Sliding window to split the data for train-validation-test. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Seizure early detection benchmark: i) Superior performance across all metrics (green [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Accuracy and recall performance for churn prediction against benchmark references. [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Average affine-invariant Riemannian distances between covariance-power features. Inter [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distilling structural signatures from churn data via GMM-based thresholding. Distinct [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Representative structural signatures extracted for seizure detection (CHB-MIT, patient [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: The main architecture ML model used. regimes—such as those arising from Mat´ern–type operators with ρ(A) bounded away from 1— these constants remain moderate, and the condition allows for many weak latent connections whose aggregate influence is “spread out” rather than concentrated along a single dominant direction. Conversely, when the condition fails, this corresponds precisely to pathological regimes i… view at source ↗
Figure 10
Figure 10. Figure 10: Performance of the core architecture (CNN+LSTM) trained on raw data for distinct [PITH_FULL_IMAGE:figures/full_fig_p028_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Performance of the core architecture (CNN+LSTM) trained with our features across [PITH_FULL_IMAGE:figures/full_fig_p029_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Performance for a stand alone CNN trained with our features across distinct patients. [PITH_FULL_IMAGE:figures/full_fig_p030_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Performance for an LSTM trained with our features across distinct patients. [PITH_FULL_IMAGE:figures/full_fig_p031_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Distribution of the learned exponent across patients. [PITH_FULL_IMAGE:figures/full_fig_p031_14.png] view at source ↗
read the original abstract

Emergent phenomena -- onset of epileptic seizures, sudden customer churn, or pandemic outbreaks -- often arise from hidden causal interactions in complex systems. We propose a machine learning method for their early detection that addresses a core challenge: unveiling and harnessing a system's latent causal structure despite the data-generating process being unknown and partially observed. The method learns an optimal feature representation from a one-parameter family of estimators -- powers of the empirical covariance or precision matrix -- offering a principled way to tune in to the underlying structure driving the emergence of critical events. A supervised learning module then classifies the learned representation. We prove structural consistency of the family and demonstrate the empirical soundness of our approach on seizure detection and churn prediction, attaining competitive results in both. Beyond prediction, and toward explainability, we ascertain that the optimal covariance power exhibits evidence of good identifiability while capturing structural signatures, thus reconciling predictive performance with interpretable statistical structure.

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 / 3 minor

Summary. The paper proposes a machine learning method for early detection of emergent phenomena (e.g., epileptic seizures, customer churn) in complex systems. It learns an optimal feature representation from the one-parameter family of powers of the empirical covariance or precision matrix, proves structural consistency of this family, applies a supervised classifier to the resulting representation, and reports competitive empirical performance on seizure and churn tasks while claiming that the optimal power exhibits good identifiability and captures structural signatures for explainability.

Significance. If the central claims hold, the work would provide a simple, tunable, and partially interpretable approach to feature construction for critical-event prediction in partially observed systems, with potential value for both accuracy and post-hoc analysis in domains such as healthcare and customer analytics. The explicit one-parameter family and the attempt to link it to structural properties distinguish it from purely black-box representations.

major comments (2)
  1. [Abstract, paragraph on core challenge and method description] Abstract and core method description: The claim that powers of the empirical covariance/precision matrix 'unveil and harness' latent causal structure under an unknown and partially observed data-generating process is not supported by a standard structural-consistency argument. Consistency of the powered estimator to a population marginal quantity does not restore missing edges, directions, or unobserved confounders; the subsequent supervised classifier may exploit any discriminative signal without demonstrating that the selected exponent has identified causal interactions rather than a convenient statistical proxy.
  2. [Method description and empirical evaluation sections] Supervised learning module and optimal-power selection: The power parameter is chosen by supervised learning on the same data used for classification. If this selection occurs via cross-validation or performance on the evaluation set, the claimed 'structural identifiability' and independence from target labels are not demonstrated; the reported performance gains may partly reflect post-hoc tuning rather than intrinsic structural recovery.
minor comments (3)
  1. [Abstract and theoretical section] The abstract asserts a proof of structural consistency yet supplies no derivation steps, assumptions on the data-generating process, or convergence rates; these should be provided in the main text or appendix with explicit conditions on partial observability.
  2. [Empirical evaluation] Empirical results lack dataset details (sample sizes, preprocessing, train/test splits), error bars, and statistical significance tests; these omissions make it difficult to assess whether the competitive results are robust.
  3. [Method] Notation for the precision matrix and its powers should be clarified to avoid ambiguity between covariance and precision versions across experiments.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help us clarify the scope of our structural consistency results and the supervised selection of the power parameter. We address each major comment below, indicating revisions where appropriate to ensure precise language without overstating causal recovery.

read point-by-point responses

  1. Referee: [Abstract, paragraph on core challenge and method description] Abstract and core method description: The claim that powers of the empirical covariance/precision matrix 'unveil and harness' latent causal structure under an unknown and partially observed data-generating process is not supported by a standard structural-consistency argument. Consistency of the powered estimator to a population marginal quantity does not restore missing edges, directions, or unobserved confounders; the subsequent supervised classifier may exploit any discriminative signal without demonstrating that the selected exponent has identified causal interactions rather than a convenient statistical proxy.

    Authors: We agree that our structural consistency result is limited to showing that the powered empirical covariance or precision matrix converges in probability to its population counterpart under standard regularity conditions on the observed data. This population quantity incorporates higher-order statistical dependencies that can serve as informative features for detecting emergent events in partially observed systems, but it does not recover unobserved confounders, edge directions, or the full causal graph. The subsequent classifier uses these features for discrimination and does not itself validate causal identification. We will revise the abstract and method sections to replace phrases such as 'unveiling and harnessing latent causal structure' with more precise wording such as 'capturing consistent statistical signatures from powered population matrices that reflect aspects of system structure', and we will add a clarifying paragraph distinguishing our approach from causal discovery methods. revision: partial

  2. Referee: [Method description and empirical evaluation sections] Supervised learning module and optimal-power selection: The power parameter is chosen by supervised learning on the same data used for classification. If this selection occurs via cross-validation or performance on the evaluation set, the claimed 'structural identifiability' and independence from target labels are not demonstrated; the reported performance gains may partly reflect post-hoc tuning rather than intrinsic structural recovery.

    Authors: The optimal power is selected via cross-validation strictly within the training folds, with the classifier trained on the chosen representation for each fold; the test set is held out entirely and used only for final evaluation. This procedure avoids leakage. We acknowledge that the selection step is supervised and therefore uses label information for optimization, so claims of complete independence from target labels require qualification. Our empirical results show that the selected power is relatively stable across independent datasets and cross-validation folds for the same type of critical event, which we interpret as evidence of identifiability in the sense of robustness rather than label-free recovery. We will revise the method and empirical sections to explicitly detail the cross-validation protocol and to rephrase 'independence from target labels' as 'the selected power exhibits stability suggestive of capturing intrinsic structural properties across instances of the phenomenon'. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on independent consistency proof

full rationale

The paper introduces a one-parameter family of matrix-power estimators and separately proves structural consistency (asymptotic convergence of the powered empirical matrix to a population counterpart). This proof is a standard large-sample result that does not invoke the supervised labels, the downstream classifier, or the particular power chosen for any finite dataset. Selection of the optimal power occurs via empirical performance on the prediction tasks, which is ordinary hyperparameter tuning rather than a claim that the consistency result itself is derived from those labels. Claims about capturing structural signatures are presented as an empirical observation that follows from the consistency guarantee and the observed performance, without any equation or step reducing the theoretical result to a fitted quantity by construction. The overall chain therefore remains self-contained against external mathematical benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that latent causal interactions exist and can be recovered from powers of second-moment statistics even when the generative process is unknown and only partially observed; the power exponent itself functions as a free parameter whose value is selected to optimize downstream classification.

free parameters (1)
  • exponent (power) of covariance or precision matrix
    Single scalar parameter that defines the family of estimators and is tuned to produce the optimal representation for the supervised task.
axioms (1)
  • domain assumption Complex systems possess latent causal structure that drives emergent critical events and is partially recoverable from second-order statistics despite unknown and incomplete observations.
    Stated in the abstract as the core challenge the method addresses.

pith-pipeline@v0.9.0 · 5699 in / 1521 out tokens · 107847 ms · 2026-05-21T14:26:15.889993+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

73 extracted references · 73 canonical work pages

  1. [1]

    What is emergence, after all?

    A. K. Rizi, “What is emergence, after all?” 2025. [Online]. Available: https://arxiv.org/abs/2507. 04951

    work page 2025

  2. [2]

    Brains, complex systems and therapeutic opportunities in epilepsy,

    R. C. Scott, “Brains, complex systems and therapeutic opportunities in epilepsy,”Seizure, vol. 90, pp. 155–159, 2021, lASSE – 15th Anniversary of Latin American Convergent Achievements in Epileptology. [Online]. Available: https://www.sciencedirect.com/science/article/ pii/S1059131121000340

    work page 2021

  3. [3]

    From the origin of life to pandemics: Emergent phenomena in complex systems,

    O. Artime and M. De Domenico, “From the origin of life to pandemics: Emergent phenomena in complex systems,”Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 380, 05 2022

    work page 2022

  4. [4]

    Epidemic processes in complex networks,

    R. Pastor-Satorras, C. Castellano, P. Van Mieghem, and A. Vespignani, “Epidemic processes in complex networks,”Rev. Mod. Phys., vol. 87, pp. 925–979, Aug 2015. [Online]. Available: https://link.aps.org/doi/10.1103/RevModPhys.87.925

    work page doi:10.1103/revmodphys.87.925 2015

  5. [5]

    Critical market crashes,

    D. Sornette, “Critical market crashes,”Physics Reports, vol. 378, no. 1, pp. 1–98, 2003. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0370157302006348

    work page 2003

  6. [6]

    Emergence in marketing: an institutional and ecosystem framework,

    S. L. Vargo, L. Peters, H. Kjellberg, K. Koskela-Huotari, S. Nenonen, F. Polese, D. Sarno, and C. Vaughan, “Emergence in marketing: an institutional and ecosystem framework,”Journal of the Academy of Marketing Science, vol. 51, no. 1, pp. 2–22, 2023

    work page 2023

  7. [7]

    Detecting emergent behavior in complex systems: A machine learning approach,

    S. S. Dahia and C. Szabo, “Detecting emergent behavior in complex systems: A machine learning approach,” inProceedings of the 38th ACM SIGSIM Conference on Principles of Advanced Discrete Simulation, ser. SIGSIM-PADS ’24. New York, NY, USA: Association for Computing Machinery, 2024, p. 81–87. [Online]. Available: https://doi.org/10.1145/3615979.3656064

    work page doi:10.1145/3615979.3656064 2024

  8. [8]

    Emergent behavior in large scale networks,

    A. Santos and J. M. F. Moura, “Emergent behavior in large scale networks,” in2011 50th IEEE Conference on Decision and Control and European Control Conference (CDC-ECC), December 2011, pp. 4485 –4490

    work page 2011

  9. [9]

    Learning the causal structure of networked dy- namical systems under latent nodes and structured noise,

    A. Santos, D. Rente, R. Seabra, and J. M. F. Moura, “Learning the causal structure of networked dy- namical systems under latent nodes and structured noise,” inProceedings of the 38th AAAI Coference on Artificial Intelligence. AAAI, 2024

    work page 2024

  10. [10]

    Propagation of transient explosive synchronization in a mesoscale mouse brain network model of epilepsy,

    A. Ranjan and S. R. Gandhi, “Propagation of transient explosive synchronization in a mesoscale mouse brain network model of epilepsy,”Network Neuroscience, vol. 8, no. 3, pp. 883–901, 10 2024. [Online]. Available: https://doi.org/10.1162/netn a 00379

    work page doi:10.1162/netn 2024

  11. [11]

    Adding connections can hinder network synchronization of time-delayed oscillators,

    J. Hart, J. Pade, T. Pereira, T. Murphy, and R. Roy, “Adding connections can hinder network synchronization of time-delayed oscillators,”Physical review. E, Statistical, nonlinear, and soft matter physics, vol. 92, p. 022804, 09 2015

    work page 2015

  12. [12]

    Pearl,Causality: Models, Reasoning, and Inference, 2nd ed

    J. Pearl,Causality: Models, Reasoning, and Inference, 2nd ed. Cambridge, UK: Cambridge University Press, 2009

    work page 2009

  13. [13]

    Smote: synthetic minority over- sampling technique,

    N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “Smote: synthetic minority over- sampling technique,”J. Artif. Int. Res., vol. 16, no. 1, p. 321–357, Jun. 2002

    work page 2002

  14. [14]

    O(n)-invariant riemannian metrics on spd matrices,

    Y. Thanwerdas and X. Pennec, “O(n)-invariant riemannian metrics on spd matrices,”Linear Algebra and its Applications, vol. 661, pp. 163–201, 2023

    work page 2023

  15. [15]

    The emergent properties of the connected brain,

    M. T. de Schotten and S. J. Forkel, “The emergent properties of the connected brain,”Science, vol. 378, no. 6619, pp. 505–510, 2022. [Online]. Available: https://www.science.org/doi/abs/10.1126/ science.abq2591

    work page 2022

  16. [16]

    The influence of inter-regional delays in generating large-scale brain networks of phase synchronization,

    N. Williams, A. Ojanper¨ a, F. Siebenh¨ uhner, B. Toselli, S. Palva, G. Arnulfo, S. Kaski, and J. Palva, “The influence of inter-regional delays in generating large-scale brain networks of phase synchronization,”NeuroImage, vol. 279, p. 120318, 2023. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S105381192300469X

    work page 2023

  17. [17]

    Latent homophily or social influence? an empirical analysis of purchase within a social network,

    L. Ma, R. Krishnan, and A. L. Montgomery, “Latent homophily or social influence? an empirical analysis of purchase within a social network,”Management Science, vol. 61, no. 2, pp. 454–473,

    work page

  18. [18]

    Available: http://dx.doi.org/10.1287/mnsc.2014.1928

    [Online]. Available: http://dx.doi.org/10.1287/mnsc.2014.1928

    work page doi:10.1287/mnsc.2014.1928 2014

  19. [19]

    Machine learning methods for gene regulatory network inference,

    A. Hegde, T. Nguyen, and J. Cheng, “Machine learning methods for gene regulatory network inference,”Briefings in Bioinformatics, vol. 26, no. 5, p. bbaf470, 09 2025. [Online]. Available: https://doi.org/10.1093/bib/bbaf470 32

    work page doi:10.1093/bib/bbaf470 2025

  20. [20]

    Social learning and distributed hypothesis testing,

    T. J. A. Lalitha and A. D. Sarwate, “Social learning and distributed hypothesis testing,”IEEE Transactions on Information Theory, vol. 64, pp. 6161–6179, September 2018

    work page 2018

  21. [21]

    Exponential collapse of social beliefs over weakly-connected heterogeneous networks,

    V. Matta, A. Santos, and A. H. Sayed, “Exponential collapse of social beliefs over weakly-connected heterogeneous networks,” inProc. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Brighton, UK, May 2019, pp. 1–5

    work page 2019

  22. [22]

    Matta, V

    V. Matta, V. Bordignon, and A. Sayed,Social Learning: Opinion Formation and Decision-Making over Graphs, 01 2025

    work page 2025

  23. [23]

    An adaptive learning approach for customer churn prediction in the telecommunication industry using evolutionary computation and na¨ ıve bayes,

    A. Amin, A. Adnan, and S. Anwar, “An adaptive learning approach for customer churn prediction in the telecommunication industry using evolutionary computation and na¨ ıve bayes,”Applied Soft Computing, vol. 137, p. 110103, 2023

    work page 2023

  24. [24]

    Customer churn prediction in the telecom sector using machine learning techniques,

    S. K. Wagh, A. A. Andhale, K. S. Wagh, J. R. Pansare, S. P. Ambadekar, and S. H. Gawande, “Customer churn prediction in the telecom sector using machine learning techniques,”Results in Control and Optimization, vol. 14, p. 100342, 2024

    work page 2024

  25. [25]

    Customer churning analysis using machine learning algorithms,

    B. Prabadevi, R. Shalini, and B. R. Kavitha, “Customer churning analysis using machine learning algorithms,”International Journal of Intelligent Networks, vol. 4, pp. 145–154, 2023

    work page 2023

  26. [26]

    New insights into churn prediction in the telecom- munication sector: A profit driven data mining approach,

    W. Verbeke, D. Martens, C. Mues, and B. Baesens, “New insights into churn prediction in the telecom- munication sector: A profit driven data mining approach,”European Journal of Operational Research, vol. 218, no. 1, pp. 211–229, 2012

    work page 2012

  27. [27]

    Churn prediction methods based on mutual customer interdependence,

    K. Ljubiˇ ci´ c, A. Mer´ cep, and Z. Kostanjcar, “Churn prediction methods based on mutual customer interdependence,”Journal of Computational Science, vol. 67, p. 101940, 03 2023

    work page 2023

  28. [28]

    Improved churn prediction in telecommuni- cation industry using data mining techniques,

    A. Keramati, R. Jafari-Marandi, M. Aliannejadiet al., “Improved churn prediction in telecommuni- cation industry using data mining techniques,”Applied Soft Computing, vol. 24, pp. 994–1012, 2014

    work page 2014

  29. [29]

    A review on machine learning methods for customer churn prediction and recommendations for business practitioners,

    A. Manzoor, M. A. Qureshi, E. Kidney, and L. Longo, “A review on machine learning methods for customer churn prediction and recommendations for business practitioners,”IEEE Access, vol. 12, pp. 70 434–70 463, 2024. [Online]. Available: https://ieeexplore.ieee.org/document/10531735

    work page arXiv 2024

  30. [30]

    A proposed hybrid framework to improve the accuracy of customer churn prediction in the telecom industry,

    S. Ouf, K. T. Mahmoud, and M. A. Abdel-Fattah, “A proposed hybrid framework to improve the accuracy of customer churn prediction in the telecom industry,”Journal of Big Data, vol. 11, no. 1, p. 70, 2024

    work page 2024

  31. [31]

    Customer churn prediction based on coordinate attention mechanism with CNN-BiLSTM,

    C. Yang, G. Xia, L. Zheng, X. Zhang, and C. Yu, “Customer churn prediction based on coordinate attention mechanism with CNN-BiLSTM,”Electronics, vol. 14, no. 10, p. 1916, 2025

    work page 1916

  32. [32]

    Customer churn prediction model based on hybrid neural networks,

    X. Liu, G. Xia, X. Zhang, W. Ma, and C. Yu, “Customer churn prediction model based on hybrid neural networks,”Scientific Reports, vol. 14, no. 1, p. 30707, 2024

    work page 2024

  33. [33]

    Integrated churn prediction and customer seg- mentation framework for telco business,

    S.-C. Wu, W.-C. Yau, T.-S. Ong, and S.-C. Chong, “Integrated churn prediction and customer seg- mentation framework for telco business,”IEEE Access, vol. 9, pp. 62 118–62 136, 2021

    work page 2021

  34. [34]

    Tempodegraphnet: Predicting user churn using dynamic social graphs and neural odes,

    M. Lee and J. Woo, “Tempodegraphnet: Predicting user churn using dynamic social graphs and neural odes,”PLoS ONE, vol. 20, no. 1, p. e0321560, 2025. [Online]. Available: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0321560

    work page doi:10.1371/journal.pone.0321560 2025

  35. [35]

    Integrated sentiment analysis and graph neural networks for understanding consumer behavior,

    D. Nimma, P. D. Sawant, S. Pokhriyal, and S. Rengaraj, “Integrated sentiment analysis and graph neural networks for understanding consumer behavior,” inProceedings of the 5th International Confer- ence on Advances in Electrical, Computing, Communication and Sustainable Technologies (ICAECT). IEEE, 2025, pp. 1–6

    work page 2025

  36. [36]

    Network neuroscience,

    D. Bassett and O. Sporns, “Network neuroscience,”Nature Neuroscience, vol. 20, pp. 353–364, 02 2017

    work page 2017

  37. [37]

    Net- work structure influences the strength of learned neural representations,

    A. E. Kahn, K. Szymula, S. Loman, E. B. Haggerty, N. Nyema, G. K. Aguirre, and D. S. Bassett, “Net- work structure influences the strength of learned neural representations,”Nature Communications, vol. 16, p. 994, 2025

    work page 2025

  38. [38]

    Disentangling the flow of signals between populations of neurons,

    E. Gokcen, A. Jasper, J. Semedo, A. Zandvakili, A. Kohn, C. Machens, and B. Yu, “Disentangling the flow of signals between populations of neurons,”Nature Computational Science, vol. 2, pp. 512–525, 08 2022

    work page 2022

  39. [39]

    Accurate identification of communication between multiple interacting neural populations,

    B. Liu, J. Sacks, and M. D. Golub, “Accurate identification of communication between multiple interacting neural populations,” inForty-second International Conference on Machine Learning,

    work page

  40. [40]

    Available: https://openreview.net/forum?id=O14GjxDAt3

    [Online]. Available: https://openreview.net/forum?id=O14GjxDAt3

    work page

  41. [41]

    Scalable and accurate deep learning with electronic health records,

    A. Rajkomar, E. Oren, K. Chen, and et al., “Scalable and accurate deep learning with electronic health records,”NPJ Digital Medicine, vol. 1, p. 18, 2018. 33

    work page 2018

  42. [42]

    Interactome networks and human disease,

    M. Vidal, M. E. Cusick, and A.-L. Barab´ asi, “Interactome networks and human disease,”Cell, vol. 144, no. 6, pp. 986–998, 2011

    work page 2011

  43. [43]

    Modern network science of neurological disorders,

    C. J. Stam, “Modern network science of neurological disorders,”Nature Reviews Neuroscience, vol. 15, no. 10, pp. 683–695, 2014

    work page 2014

  44. [44]

    Seizure prediction: the long and winding road,

    F. Mormann, R. G. Andrzejak, C. E. Elger, and K. Lehnertz, “Seizure prediction: the long and winding road,”Brain, vol. 130, no. 2, pp. 314–333, 2007

    work page 2007

  45. [45]

    A novel epileptic seizure detection system using scalp EEG signals based on hybrid CNN-SVM classifier,

    A. Saidi, S. Ben Othman, and S. Ben Saoud, “A novel epileptic seizure detection system using scalp EEG signals based on hybrid CNN-SVM classifier,” inProceedings of the 2021 IEEE Symposium on Industrial Electronics & Applications (ISIEA). IEEE, 2021

    work page 2021

  46. [46]

    Progressive Fourier Transform (PFT): Enhancing time-frequency representation of EEG signals for stress and seizure detection,

    N. S. Amer, S. B. Belhaouari, and H. Bensmail, “Progressive Fourier Transform (PFT): Enhancing time-frequency representation of EEG signals for stress and seizure detection,” inProceedings of the 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2023, pp. 2441–2448

    work page 2023

  47. [47]

    Detecting epileptic seizures in EEG signals with complementary ensemble empirical mode decomposition and extreme gradient boosting,

    J. Wu, T. Zhou, and T. Li, “Detecting epileptic seizures in EEG signals with complementary ensemble empirical mode decomposition and extreme gradient boosting,”Entropy, vol. 22, no. 2, p. 140, 2020

    work page 2020

  48. [48]

    Epileptic disorder detection of seizures using EEG signals,

    M. K. Alharthi, K. M. Moria, D. M. Alghazzawi, and H. O. Tayeb, “Epileptic disorder detection of seizures using EEG signals,”Sensors, vol. 22, no. 17, p. 6592, 2022

    work page 2022

  49. [49]

    A deep learning approach for automatic seizure detection in children with epilepsy,

    A. Abdelhameed and B. Magdy, “A deep learning approach for automatic seizure detection in children with epilepsy,”Frontiers in Computational Neuroscience, vol. 15, p. 645610, 2021. [Online]. Available: https://www.frontiersin.org/articles/10.3389/fncom.2021.645610/full

    work page doi:10.3389/fncom.2021.645610/full 2021

  50. [50]

    One-dimensional convolutional neural networks for seizure onset detection using long-term scalp and intracranial EEG,

    X. Wang, X. Wang, W. Liu, Z. Chang, T. K¨ arkk¨ ainen, and F. Cong, “One-dimensional convolutional neural networks for seizure onset detection using long-term scalp and intracranial EEG,”Neurocom- puting, vol. 459, pp. 212–222, 2021

    work page 2021

  51. [51]

    Epileptic seizure detection based on bidirectional gated recurrent unit network,

    Y. Zhang, S. Yao, R. Yang, X. Liu, W. Qiu, L. Han, W. Zhou, and W. Shang, “Epileptic seizure detection based on bidirectional gated recurrent unit network,”IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 30, pp. 135–145, 2022

    work page 2022

  52. [52]

    Multi-dimensional hybrid bilinear CNN-LSTM models for epileptic seizure detection and prediction using EEG signals,

    S. Liu, J. Wang, S. Li, and L. Cai, “Multi-dimensional hybrid bilinear CNN-LSTM models for epileptic seizure detection and prediction using EEG signals,”Journal of Neural Engineering, vol. 21, no. 6, p. 066045, 2024

    work page 2024

  53. [53]

    Seizure prediction using directed transfer function and convolution neural network on intracranial EEG,

    G. Wang, D. Wang, C. Du, K. Li, J. Zhang, Z. Liu, Y. Tao, M. Wang, Z. Cao, and X. Yan, “Seizure prediction using directed transfer function and convolution neural network on intracranial EEG,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 28, no. 12, pp. 2711–2720, 2020

    work page 2020

  54. [54]

    Convolutional neural networks for epileptic seizure detec- tion using eeg signals,

    N. Truong, A. Nguyen, and T. A. Khoa, “Convolutional neural networks for epileptic seizure detec- tion using eeg signals,” in2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2018, pp. 1997–2001

    work page 2018

  55. [55]

    Deep learning-based electroencephalogram analysis for epileptic seizure detection: A review,

    T. Liu, C. Zhang, J. Wang, and J. Li, “Deep learning-based electroencephalogram analysis for epileptic seizure detection: A review,”IEEE Access, vol. 8, pp. 113 909–113 923, 2020

    work page 2020

  56. [56]

    Temporal graph convolutional networks for automatic seizure detection,

    I. Covert, J. Sun, and Y. Halpern, “Temporal graph convolutional networks for automatic seizure detection,” inProceedings of the Machine Learning for Healthcare Conference, ser. Proceedings of Machine Learning Research, vol. 106. PMLR, 2019, pp. 160–176

    work page 2019

  57. [57]

    EEG seizure prediction with graph neural networks,

    F. Yanget al., “EEG seizure prediction with graph neural networks,”Neural Networks, vol. 145, pp. 199–210, 2022

    work page 2022

  58. [58]

    Efficient graph convolutional networks for seizure prediction using EEG signals,

    A. Razi, R. Hussein, D. Nhu, Q. Wang, Z. Zhou, and R. Cipolla, “Efficient graph convolutional networks for seizure prediction using EEG signals,”Frontiers in Neuroscience, vol. 16, p. 967116, 2022

    work page 2022

  59. [59]

    Epileptic seizure prediction using stacked CNN- BiLSTM: A novel approach,

    Z. F. Quadri, M. S. Akhoon, and S. A. Loan, “Epileptic seizure prediction using stacked CNN- BiLSTM: A novel approach,”IEEE Transactions on Artificial Intelligence, 2024

    work page 2024

  60. [60]

    Automated seizure detection in epilepsy using a novel dynamic temporal-spatial graph attention network,

    K. Yan, X. Luo, L. Ye, W. Geng, J. He, I. N. Mu, X. Hou, X. Zan, J. Ma, F. Li, L. Zhang, and X. Chou, “Automated seizure detection in epilepsy using a novel dynamic temporal-spatial graph attention network,”Scientific Reports, vol. 15, 05 2025

    work page 2025

  61. [61]

    Self-supervised graph neural networks for improved electroencephalographic seizure analysis,

    S. Tang, J. Dunnmon, K. K. Saab, X. Zhang, Q. Huang, F. Dubost, D. Rubin, and C. Lee-Messer, “Self-supervised graph neural networks for improved electroencephalographic seizure analysis,” inThe Tenth International Conference on Learning Representations, ser. ICLR’22. JMLR.org, 2022. 34

    work page 2022

  62. [62]

    Rest: efficient and accelerated eeg seizure analysis through residual state updates,

    A. Afzal, G. Chrysos, V. Cevher, and M. Shoaran, “Rest: efficient and accelerated eeg seizure analysis through residual state updates,” inProceedings of the 41st International Conference on Machine Learning, ser. ICML’24. JMLR.org, 2024

    work page 2024

  63. [63]

    Mat´ ern gaussian processes on graphs,

    V. Borovitskiy, A. Terenin, P. Mostowsky, and M. Deisenroth, “Mat´ ern gaussian processes on graphs,” Advances in Neural Information Processing Systems, vol. 34, pp. 14 507–14 519, 2021

    work page 2021

  64. [64]

    An explicit link between Gaussian fields and Gaussian markov random fields: the spde approach,

    F. Lindgren, H. Rue, and J. Lindstr¨ om, “An explicit link between Gaussian fields and Gaussian markov random fields: the spde approach,”Journal of the Royal Statistical Society: Series B, vol. 73, no. 4, pp. 423–498, 2011

    work page 2011

  65. [65]

    Bayesian generalized linear modeling of cortical surface fMRI,

    A. F. Mejia, Y. Yue, D. Bolin, F. Lindgren, and J. E. Taylor, “Bayesian generalized linear modeling of cortical surface fMRI,”NeuroImage, vol. 218, p. 116935, 2020

    work page 2020

  66. [66]

    Spatial Bayesian GLM on cortical surfaces for fMRI data,

    J. Spencer, A. F. Mejia, and J. Taylor, “Spatial Bayesian GLM on cortical surfaces for fMRI data,” NeuroImage, vol. 257, p. 119316, 2022

    work page 2022

  67. [67]

    Fast Bayesian whole-brain fMRI analysis with a 3D spatial model,

    P. Sid´ en, A. Eklund, D. Bolin, and F. Lindgren, “Fast Bayesian whole-brain fMRI analysis with a 3D spatial model,”NeuroImage, vol. 242, p. 118456, 2021

    work page 2021

  68. [68]

    The SPDE approach for Gaussian and non-Gaussian fields: ten years later,

    F. Lindgren, D. Bolin, and H. Rue, “The SPDE approach for Gaussian and non-Gaussian fields: ten years later,”Spatial Statistics, 2024, in press

    work page 2024

  69. [69]

    Revisiting correlation-based func- tional connectivity and its relationship with structural connectivity,

    R. Li´ egeois, A. Santos, V. Matta, D. Van De Ville, and A. H. Sayed, “Revisiting correlation-based func- tional connectivity and its relationship with structural connectivity,”Network Neuroscience, vol. 4, no. 4, pp. 1235–1251, 2020

    work page 2020

  70. [70]

    Koller and N

    D. Koller and N. Friedman,Probabilistic Graphical Models: Principles and Techniques - Adaptive Computation and Machine Learning. The MIT Press, 2009

    work page 2009

  71. [71]

    Which graphical models are difficult to learn?

    A. Montanari and J. Pereira, “Which graphical models are difficult to learn?” inAdvances in Neural Information Processing Systems, vol. 22, Vancouver, Canada, 2009

    work page 2009

  72. [72]

    R. A. Horn and C. R. Johnson,Matrix Analysis, 2nd ed. New York, NY, USA: Cambridge University Press, 2012

    work page 2012

  73. [73]

    Kato,Perturbation Theory for Linear Operators, 2nd ed., ser

    T. Kato,Perturbation Theory for Linear Operators, 2nd ed., ser. Classics in Mathematics. Berlin Heidelberg: Springer-Verlag, 1995. 35

    work page 1995