pith. sign in

arxiv: 2605.21192 · v1 · pith:74ZORLBSnew · submitted 2026-05-20 · 💻 cs.CE · q-fin.CP

The Statistical Significance of the Inclusion of Graph Neural Networks in the Financial Time Series Forecasting Problem

Marco Gregnanin , Johannes De Smedt , Giorgio Gnecco , Maurizio Parton This is my paper

Pith reviewed 2026-05-21 01:23 UTC · model grok-4.3

classification 💻 cs.CE q-fin.CP
keywords graph neural networksfinancial time seriesforecastinggeometric patternsstatistical significanceunivariate time series
0
0 comments X

The pith

Including Graph Neural Networks to capture geometric patterns leads to statistically significant improvements in financial time series forecasting.

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

The paper examines whether adding geometric information from Graph Neural Networks helps predict financial time series better than models that only look at time patterns. It proposes the Time-Geometric model that combines both types of patterns. Extensive tests show that the added geometric patterns produce forecasting gains that pass statistical significance checks. A sympathetic reader would care because financial forecasting is hard and small improvements can matter for decisions. This suggests that time series data contain useful structure beyond pure sequences.

Core claim

The authors introduce the Time-Geometric model as a combination of models that exploit both geometric and temporal patterns in univariate financial time series. Through empirical evaluations, they demonstrate that leveraging geometric patterns captured through Graph Neural Networks yields statistically significant improvements in forecasting accuracy over models relying solely on temporal patterns.

What carries the argument

The Time-Geometric model, which integrates Graph Neural Networks to extract geometric patterns in addition to standard temporal analysis.

If this is right

  • Standard temporal models can be enhanced by incorporating geometric patterns from GNNs.
  • The improvements in accuracy are statistically significant rather than due to chance.
  • Geometric patterns provide complementary information to temporal patterns in financial data.
  • Extensive empirical evaluations support the inclusion of such geometric components.

Where Pith is reading between the lines

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

  • This could encourage exploring graph-based representations for other types of sequential data beyond finance.
  • Researchers might investigate how to best construct the graphs that GNNs operate on for time series.
  • Similar combinations could be tested in non-financial domains like energy consumption or traffic flow prediction.

Load-bearing premise

The geometric patterns extracted by the GNN provide information that is independent from the temporal patterns used by standard models.

What would settle it

Running the same experiments on the datasets used and finding that the combined model does not show statistically significant improvement over the temporal-only baseline.

Figures

Figures reproduced from arXiv: 2605.21192 by Giorgio Gnecco, Johannes De Smedt, Marco Gregnanin, Maurizio Parton.

Figure 1
Figure 1. Figure 1: illustrates the general architecture of the model [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The Visibility Graph algorithm was applied to a time series generated from a Geometric Brownian Motion with the initial value equal to 100 and with mean and variance equal to 0.05 and 0.5, respectively. The computation of the algorithm was conducted using the “time series to visibility graphs” (ts2vg) Python package (Bergillos, 2020). 4.2. Time-Geometric Model The Time-Geometric model receives input in the… view at source ↗
Figure 3
Figure 3. Figure 3: Time-Geometric Model Architecture. The baseline model differs in the absence of both the input At and the Ge￾ometric Component. TIME COMPONENT The Time Component takes the feature matrix Xt ∈ R m×F ′ as input and produces the output X˜ T IME t ∈ R m×F ′ . Its objective is to analyze and leverage temporal patterns within the data. Algorithm 1 outlines the operations of this compo￾nent. Algorithm 1 Time Comp… view at source ↗
Figure 4
Figure 4. Figure 4: Average metric values for 1 day prediction for all the models within the considered dataset are presented in the figure. The x-axis denotes the models used both as baselines and in the Time Component. The blue bar represents the average metric values for the respective baseline model, while the red bar signifies the average metric value for the Time-Geometric model [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Namenyi test for 1 day forecasting. Non-statistically significant relationships are highlighted in yellow. analysis for the 5 day and 20 day predictions is detailed in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Example of degree distribution for a regular graph, a random graph, and a small-world graph. dataset containing M elements, these evaluation metrics are defined as follows: RMSE = vuut 1 M X M i=1 (yi − yˆi) 2 , MAE = 1 M X M i=1 |yi − yˆi | , MAP E = 1 M X M i=1 [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Average metric values for 5 day prediction for all the models within the considered dataset are presented in the figure. The x-axis denotes the models used both as baselines and in the Time Component. The blue bar represents the average metric values for the respective baseline model, while the red bar signifies the average metric value for the Time-Geometric model. E. Statistical Significance In this sect… view at source ↗
Figure 8
Figure 8. Figure 8: Average metric values for 20 day prediction for all the models within the considered dataset are presented in the figure. The x-axis denotes the models used both as baselines and in the Time Component. The blue bar represents the average metric values for the respective baseline model, while the red bar signifies the average metric value for the Time-Geometric model. all models for each dataset, assigning … view at source ↗
Figure 9
Figure 9. Figure 9: Namenyi test for 5 day forecasting. Non-statistically significant relationship are highlighted in yellow. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Namenyi test for 20 day forecasting. Non-statistically significant relationship are highlighted in yellow. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_10.png] view at source ↗
read the original abstract

Forecasting univariate time series in the financial market is a challenging endeavor. While numerous statistical and machine learning models have been introduced to address this challenge, they typically concentrate solely on analyzing temporal patterns within the time series data. In this research, we study the statistical significance of the inclusion of geometric patterns in enhancing forecasting accuracy within the context of time series analysis. We introduce the Time-Geometric model, a combination of models designed to exploit both geometric and temporal patterns. The contribution of this research lies in advancing the domain of univariate time series prediction,as demonstrated through extensive empirical evaluations. Our findings underscore that leveraging geometric patterns, captured through Graph Neural Networks, yields statistically significant improvements in forecasting accuracy.

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

1 major / 1 minor

Summary. The manuscript claims that a Time-Geometric model, which combines temporal patterns with geometric patterns captured via Graph Neural Networks, produces statistically significant improvements in accuracy for univariate financial time series forecasting, as shown by extensive empirical evaluations.

Significance. If the empirical results hold, the work would be significant for financial time series forecasting by indicating that GNN-derived geometric patterns supply predictive information beyond standard temporal models.

major comments (1)
  1. [Abstract] Abstract: The central claim that leveraging geometric patterns captured through Graph Neural Networks yields statistically significant improvements is unsupported by any experimental details. The abstract supplies no information on the temporal baseline models, datasets or assets, forecasting horizons, integration method for the GNN component, loss functions, evaluation metrics, or the statistical tests (including multiplicity corrections) used to establish significance. This prevents assessment of whether the geometric patterns add independent value or whether the reported significance is valid.
minor comments (1)
  1. [Abstract] Abstract: missing space after comma in 'prediction,as demonstrated'.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback. We address the concern regarding the abstract below and agree that revisions are needed to strengthen the presentation of our claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that leveraging geometric patterns captured through Graph Neural Networks yields statistically significant improvements is unsupported by any experimental details. The abstract supplies no information on the temporal baseline models, datasets or assets, forecasting horizons, integration method for the GNN component, loss functions, evaluation metrics, or the statistical tests (including multiplicity corrections) used to establish significance. This prevents assessment of whether the geometric patterns add independent value or whether the reported significance is valid.

    Authors: We agree that the abstract is currently too concise and omits key experimental details, which limits the ability to assess the contribution of the geometric patterns and the validity of the reported significance. In the revised manuscript we will expand the abstract to include brief but specific information on the temporal baseline models, the financial datasets and assets examined, the forecasting horizons, the method used to integrate the GNN component with the temporal models, the loss functions and evaluation metrics, and the statistical tests (including any multiplicity corrections). These additions will make the central claim more transparent and allow readers to better evaluate whether the GNN-derived geometric patterns supply independent predictive value. revision: yes

Circularity Check

0 steps flagged

No circularity: abstract states empirical claim with no derivation or equations

full rationale

The available text consists solely of the abstract, which presents the work as an empirical study introducing a Time-Geometric model that combines temporal and geometric patterns (via GNNs) and reports statistically significant forecasting improvements from extensive evaluations. No equations, first-principles derivations, fitted parameters renamed as predictions, self-citations, or ansatzes are present. The central claim is framed as a data-driven finding rather than a mathematical reduction, so no load-bearing step reduces to its own inputs by construction. The derivation chain is therefore self-contained and exhibits no circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The central claim rests on the effectiveness of the newly introduced Time-Geometric model and on the validity of the statistical significance tests; the abstract provides no explicit free parameters, background axioms, or independent evidence for the model.

invented entities (1)
  • Time-Geometric model no independent evidence
    purpose: Combination of geometric (GNN) and temporal pattern models for univariate time series forecasting
    Introduced in the abstract as the core contribution that exploits both geometric and temporal patterns.

pith-pipeline@v0.9.0 · 5625 in / 1146 out tokens · 43513 ms · 2026-05-21T01:23:28.286515+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages · 3 internal anchors

  1. [1]

    Proceedings of the National Academy of Sciences , volume=

    From time series to complex networks: The visibility graph , author=. Proceedings of the National Academy of Sciences , volume=. 2008 , publisher=

    work page 2008

  2. [2]

    Ts2vg: Time series to visibility graphs

    Carlos Bergillos , howpublished = ". Ts2vg: Time series to visibility graphs

    work page

  3. [3]

    PloS One , volume=

    Visibility Graph Based Time Series Analysis , author=. PloS One , volume=. 2015 , publisher=

    work page 2015

  4. [4]

    2016 , publisher=

    Bishop, Christopher M , title=. 2016 , publisher=

    work page 2016

  5. [5]

    2005 , publisher=

    Analysis of Financial Time Series , author=. 2005 , publisher=

    work page 2005

  6. [6]

    2011 , publisher=

    Introduction to Stochastic Calculus Applied to Finance , author=. 2011 , publisher=

    work page 2011

  7. [7]

    2015 , publisher=

    Box, George EP and Jenkins, Gwilym M and Reinsel, Gregory C and Ljung, Greta M , title=. 2015 , publisher=

    work page 2015

  8. [8]

    Journal of Political Economy , volume=

    The pricing of options and corporate liabilities , author=. Journal of Political Economy , volume=. 1973 , publisher=

    work page 1973

  9. [9]

    2003 , publisher=

    Financial Modelling with Jump Processes , author=. 2003 , publisher=

    work page 2003

  10. [10]

    Cognitive Science , volume=

    Finding structure in time , author=. Cognitive Science , volume=. 1990 , publisher=

    work page 1990

  11. [11]

    IEEE Communications Magazine , volume=

    Deep learning with long short-term memory for time series prediction , author=. IEEE Communications Magazine , volume=. 2019 , publisher=

    work page 2019

  12. [12]

    Deep Learning , author=

    work page

  13. [13]

    Advances in Neural Information Processing Systems , volume=

    Attention is all you need , author=. Advances in Neural Information Processing Systems , volume=

    work page

  14. [14]

    2016 , eprint=

    Temporal Convolutional Networks: A Unified Approach to Action Segmentation , author=. 2016 , eprint=

    work page 2016

  15. [15]

    Selecta volume E , author=

    Fractals and Scaling in Finance: Discontinuity, Concentration, Risk. Selecta volume E , author=. 2013 , publisher=

    work page 2013

  16. [16]

    1994 , publisher=

    Fractal Market Analysis: Applying Chaos Theory to Investment and Economics , author=. 1994 , publisher=

    work page 1994

  17. [17]

    Fractals , volume=

    Fractal Geometry of Financial Time Series , author=. Fractals , volume=. 1995 , publisher=

    work page 1995

  18. [18]

    IEEE Transactions on Neural Networks , volume=

    The graph neural network model , author=. IEEE Transactions on Neural Networks , volume=. 2008 , publisher=

    work page 2008

  19. [19]

    Advances in Neural Information Processing Systems , volume=

    Inductive representation learning on large graphs , author=. Advances in Neural Information Processing Systems , volume=

    work page

  20. [20]

    Neural Computation , volume=

    Long short-term memory , author=. Neural Computation , volume=. 1997 , publisher=

    work page 1997

  21. [21]

    Wu, Zonghan and Pan, Shirui and Chen, Fengwen and Long, Guodong and Zhang, Chengqi and Yu, Philip S. , year=. A Comprehensive Survey on Graph Neural Networks , volume=. IEEE Transactions on Neural Networks and Learning Systems , publisher=. doi:10.1109/tnnls.2020.2978386 , number=

    work page doi:10.1109/tnnls.2020.2978386 2020

  22. [22]

    Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting , url=

    Yu, Bing and Yin, Haoteng and Zhu, Zhanxing , year=. Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting , url=. doi:10.24963/ijcai.2018/505 , booktitle=

    work page doi:10.24963/ijcai.2018/505 2018

  23. [23]

    Advances in Neural Information Processing Systems , volume=

    Spectral temporal graph neural network for multivariate time-series forecasting , author=. Advances in Neural Information Processing Systems , volume=

    work page

  24. [24]

    IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 21, 3848–3858

    Zhao, Ling and Song, Yujiao and Zhang, Chao and Liu, Yu and Wang, Pu and Lin, Tao and Deng, Min and Li, Haifeng , year=. IEEE Transactions on Intelligent Transportation Systems , publisher=. doi:10.1109/tits.2019.2935152 , number=

    work page doi:10.1109/tits.2019.2935152 2019

  25. [25]

    International Conference on Learning Representations , year=

    Graph Wavelet Neural Network , author=. International Conference on Learning Representations , year=

    work page

  26. [26]

    IEEE Transactions on Signal Processing , volume=

    Gated graph recurrent neural networks , author=. IEEE Transactions on Signal Processing , volume=. 2020 , publisher=

    work page 2020

  27. [27]

    2022 , issn =

    Financial time series forecasting with multi-modality graph neural network , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.patcog.2021.108218 , url =

    work page doi:10.1016/j.patcog.2021.108218 2022

  28. [28]

    Resources Policy , volume =

    A new hybrid model for multi-step. Resources Policy , volume =. 2023 , issn =. doi:https://doi.org/10.1016/j.resourpol.2023.103956 , url =

    work page doi:10.1016/j.resourpol.2023.103956 2023

  29. [29]

    2022 , isbn =

    Xiang, Sheng and Cheng, Dawei and Shang, Chencheng and Zhang, Ying and Liang, Yuqi , title =. 2022 , isbn =. doi:10.1145/3511808.3557089 , booktitle =

    work page doi:10.1145/3511808.3557089 2022

  30. [30]

    Mathematics , volume=

    A Combined Model Based on Recurrent Neural Networks and Graph Convolutional Networks for Financial Time Series Forecasting , author=. Mathematics , volume=. 2023 , publisher=

    work page 2023

  31. [31]

    Machine Learning with Applications , volume=

    Rainfall prediction: A comparative analysis of modern machine learning algorithms for time-series forecasting , author=. Machine Learning with Applications , volume=. 2022 , publisher=

    work page 2022

  32. [32]

    Journal of Healthcare Engineering , year=

    Comparison of Time Series Methods and Machine Learning Algorithms for Forecasting Taiwan Blood Services Foundation's Blood Supply , author=. Journal of Healthcare Engineering , year=

    work page

  33. [33]

    2019 , issn =

    Evaluation of statistical and machine learning models for time series prediction: Identifying the state-of-the-art and the best conditions for the use of each model , journal =. 2019 , issn =. doi:https://doi.org/10.1016/j.ins.2019.01.076 , url =

    work page doi:10.1016/j.ins.2019.01.076 2019

  34. [34]

    The Journal of Machine Learning Research , volume=

    Statistical comparisons of classifiers over multiple data sets , author=. The Journal of Machine Learning Research , volume=. 2006 , publisher=

    work page 2006

  35. [35]

    Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling

    Empirical evaluation of gated recurrent neural networks on sequence modeling , author=. arXiv preprint arXiv:1412.3555 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  36. [36]

    IEEE Transactions on Signal Processing , volume=

    Bidirectional recurrent neural networks , author=. IEEE Transactions on Signal Processing , volume=. 1997 , publisher=

    work page 1997

  37. [37]

    Bidirectional

    Graves, Alex and Fern. Bidirectional. International Conference on Artificial Neural Networks , pages=. 2005 , organization=

    work page 2005

  38. [38]

    International Conference on Machine Learning , pages=

    Dynamic memory networks for visual and textual question answering , author=. International Conference on Machine Learning , pages=. 2016 , organization=

    work page 2016

  39. [39]

    Semi-Supervised Classification with Graph Convolutional Networks

    Semi-supervised classification with graph convolutional networks , author=. arXiv preprint arXiv:1609.02907 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  40. [40]

    Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

    Optuna: A next-generation hyperparameter optimization framework , author=. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

    work page

  41. [41]

    International Journal of Forecasting , volume=

    Another look at measures of forecast accuracy , author=. International Journal of Forecasting , volume=. 2006 , publisher=

    work page 2006

  42. [42]

    Adam: A Method for Stochastic Optimization

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  43. [43]

    Journal of the American Statistical Association , volume=

    The use of ranks to avoid the assumption of normality implicit in the analysis of variance , author=. Journal of the American Statistical Association , volume=. 1937 , publisher=

    work page 1937

  44. [44]

    The Annals of Mathematical Statistics , volume=

    A comparison of alternative tests of significance for the problem of m rankings , author=. The Annals of Mathematical Statistics , volume=. 1940 , publisher=

    work page 1940

  45. [45]

    1963 , publisher=

    Distribution-Free Multiple Comparisons , author=. 1963 , publisher=

    work page 1963

  46. [46]

    Data Mining and Knowledge Discovery , volume=

    On comparing classifiers: Pitfalls to avoid and a recommended approach , author=. Data Mining and Knowledge Discovery , volume=. 1997 , publisher=

    work page 1997

  47. [47]

    2003 , publisher=

    Handbook of Parametric and Nonparametric Statistical Procedures , author=. 2003 , publisher=

    work page 2003

  48. [48]

    Breakthroughs in Statistics: Methodology and Distribution , pages=

    Individual comparisons by ranking methods , author=. Breakthroughs in Statistics: Methodology and Distribution , pages=. 1992 , publisher=

    work page 1992

  49. [49]

    ACM Computing Surveys , volume=

    Deep learning for time series forecasting: Tutorial and literature survey , author=. ACM Computing Surveys , volume=. 2022 , publisher=

    work page 2022

  50. [50]

    Machine Learning and Big Data Analytics Paradigms: Analysis, Applications and Challenges , pages=

    A survey on deep learning for time-series forecasting , author=. Machine Learning and Big Data Analytics Paradigms: Analysis, Applications and Challenges , pages=. 2021 , publisher=

    work page 2021

  51. [51]

    2016 International Conference on Computing Technologies and Intelligent Data Engineering (ICCTIDE'16) , pages=

    A survey on forecasting of time series data , author=. 2016 International Conference on Computing Technologies and Intelligent Data Engineering (ICCTIDE'16) , pages=. 2016 , organization=

    work page 2016

  52. [52]

    1991 , publisher=

    Time Series: Theory and Methods , author=. 1991 , publisher=

    work page 1991

  53. [53]

    Investigation of market efficiency and financial stability between

    Rounaghi, Mohammad Mahdi and Zadeh, Farzaneh Nassir , journal=. Investigation of market efficiency and financial stability between. 2016 , publisher=

    work page 2016

  54. [54]

    2020 , publisher=

    Time Series Analysis , author=. 2020 , publisher=

    work page 2020

  55. [55]

    Stock price prediction using the

    Ariyo, Adebiyi A and Adewumi, Adewumi O and Ayo, Charles K , booktitle=. Stock price prediction using the. 2014 , organization=

    work page 2014

  56. [56]

    Econometrica: Journal of the Econometric Society , pages=

    Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation , author=. Econometrica: Journal of the Econometric Society , pages=. 1982 , publisher=

    work page 1982

  57. [57]

    Comparative study of volatility forecasting models: The case of

    Lee, San K and Nguyen, LT and Sy, Malick O , journal=. Comparative study of volatility forecasting models: The case of

    work page

  58. [58]

    1994 , publisher=

    Bollerslev, Tim and Engle, Robert F and Nelson, Daniel B , journal=. 1994 , publisher=

    work page 1994

  59. [59]

    Francq, Christian and Zakoian, Jean-Michel , year=

    work page

  60. [60]

    2016 IEEE International Conference on Data Science and Advanced Analytics (DSAA) , pages=

    Robust online time series prediction with recurrent neural networks , author=. 2016 IEEE International Conference on Data Science and Advanced Analytics (DSAA) , pages=. 2016 , organization=

    work page 2016

  61. [61]

    Stock price prediction using

    Ghosh, Achyut and Bose, Soumik and Maji, Giridhar and Debnath, Narayan and Sen, Soumya , booktitle=. Stock price prediction using

    work page

  62. [62]

    Research on Improved

    Chen, Chi and Xue, Lei and Xing, Wanqi , journal=. Research on Improved. 2023 , publisher=

    work page 2023

  63. [63]

    Stock price prediction using

    Selvin, Sreelekshmy and Vinayakumar, R and Gopalakrishnan, EA and Menon, Vijay Krishna and Soman, KP , booktitle=. Stock price prediction using. 2017 , organization=

    work page 2017

  64. [64]

    Predicting stock prices using

    Roondiwala, Murtaza and Patel, Harshal and Varma, Shraddha , journal=. Predicting stock prices using

    work page

  65. [65]

    Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages=

    Stock price prediction via discovering multi-frequency trading patterns , author=. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages=

    work page

  66. [66]

    Hierarchical Multi-Scale

    Ding, Qianggang and Wu, Sifan and Sun, Hao and Guo, Jiadong and Guo, Jian , booktitle =. Hierarchical Multi-Scale. 2020 , month =. doi:10.24963/ijcai.2020/640 , url =

    work page doi:10.24963/ijcai.2020/640 2020

  67. [67]

    Lu, Wenjie and Li, Jiazheng and Wang, Jingyang and Qin, Lele , journal=. A. 2021 , publisher=

    work page 2021

  68. [68]

    A new attention-based

    Lin, Yuyang and Huang, Qi and Zhong, Qiyin and Li, Muyang and Li, Yan and Ma, Fei , journal=. A new attention-based. 2022 , publisher=

    work page 2022

  69. [69]

    Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume=

    Network science , author=. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume=. 2013 , publisher=

    work page 2013

  70. [70]

    Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    ROLAND: graph learning framework for dynamic graphs , author=. Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    work page