Pith. sign in

REVIEW 4 major objections 6 minor 73 references

Community-Aware Temporal Walks: Parameter-Free Representation Learning on Continuous-Time Dynamic Graphs

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read CTWalks claims that guiding temporal walks by community boundaries, tagging them with community labels, and encoding them with ODEs yields more accurate temporal link prediction than six established baselines.

desk verdict CTWalks is a plausible new method with informative ablations, but the 27x discrepancy in the Taobao dataset between Table 1 and Appendix D.1 makes the headline empirical claims untrustworthy until resolved. read the letter →

arxiv 2501.11880 v1 pith:6YER3VEW submitted 2025-01-21 cs.LG cs.AI

He Yu , Jing Liu This is my paper · ORCID
classification cs.LGcs.AI
keywords continuous-timedynamicgraphstemporallinkpredictionwalkscommunitydetectionanonymizedwalkencodingneuralODEinductivelearningparameter-freesampling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

CTWalks is a representation-learning method for continuous-time dynamic graphs, where edges are time-stamped interactions rather than snapshots. It builds a weighted static graph from interaction counts, detects communities on it, and then samples temporal walks that stay inside a community for ordinary nodes and run between communities only through bridging nodes, so no extra parameter decides exploration versus exploitation. Each sampled walk is anonymized by position counts and tagged with the community labels of the two root nodes, and the sequence is encoded by alternating GRU-style updates with an ODE that integrates the hidden state over irregular time gaps. The paper argues that community-aware sampling breaks locality bias, that the encoding corresponds to a low-rank matrix factorization, and that the full pipeline outperforms six established baselines on temporal link prediction across five datasets, with the largest margins in inductive settings involving unseen nodes. If these results hold, CTWalks offers a way to inject mesoscopic structure into dynamic graph embeddings without hand-tuned walk parameters.

What carries the argument

The load-bearing object is the community-restricted temporal walk with its two-layer transition structure. A weighted temporal graph $G_w$ is formed by setting edge weights to interaction counts, then modularity optimization partitions nodes into communities; bridging nodes belong to multiple communities, while non-bridging nodes are confined to one. During sampling, a walk from a non-bridging node can only follow edges inside its community, while a walk from a bridging node can only follow edges in the inter-community subgraph, with transition probability proportional to $e^{-(t-t')}$ over valid neighbors. The anonymization step replaces node identities with position-based occurrence counts and appends the community labels of the source and target roots, restoring context that purely structural anonymization loses. The encoding alternates an instantaneous GRU update $g$ with a continuous ODE integration $f$ over each inter-event interval, so the final hidden state carries both discrete structural updates and continuous temporal evolution.

What would settle it

Compare CTWalks with an unrestricted temporal-walk variant on a dataset whose training-time communities are weak or whose future edges mostly join nodes from different communities; if the unrestricted variant matches or exceeds CTWalks on AUC, the community-restriction claim is falsified.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that community structure, treated as a mesoscopic layer between local neighborhoods and the whole graph, can be made the organizing principle for temporal walk sampling, anonymization, and continuous-time encoding. The method partitions the aggregated interaction graph into communities via modularity optimization, classifies nodes as bridging or non-bridging, restricts each walk to the appropriate intra- or inter-community subgraph, and then encodes the resulting anonymized walks with an ODE-based continuous integration step. The paper claims a theoretical lemma showing this restriction reduces locality bias relative to uniform random walks, and a second lemma showing the learned embeddings factor a shifted pointwise mutual information matrix built from intra- and inter-community transition matrices. Empirically, it reports that CTWalks achieves the best or second-best AUC and AP against six baselines on five datasets, especially for inductive link prediction with new nodes.

Load-bearing premise

The method assumes that the community structure computed once from the training period stays meaningful for future edges, so that forbidding walks that cross community boundaries except through bridging nodes does not remove the very patterns needed to predict which nodes will connect later.

Editorial extensions

If this is right

  • If CTWalks is correct, temporal link prediction on unseen nodes should benefit most, since anonymized positional encodings plus community labels let the model transfer across nodes with similar roles.
  • If CTWalks is correct, sampling no longer needs hand-tuned walk-direction hyperparameters: community membership alone decides whether a walk explores inside a community or across bridging nodes.
  • If CTWalks is correct, irregular interaction times can be handled without discretization or imputation, because the hidden state evolves continuously between observed events.
  • If CTWalks is correct, the learned embeddings admit a matrix-factorization interpretation, giving a principled account of what the community-restricted walks encode.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: if community restriction is the real source of gains, CTWalks should degrade on graphs with low modularity or heavy cross-community churn; a concrete test is to compare CTWalks against an unrestricted temporal-walk variant on such graphs.
  • Beyond the paper: the matrix-factorization view suggests CTWalks inherits Skip-Gram with negative sampling biases, so high-degree bridging nodes may dominate the embedding; measuring predictive accuracy stratified by node degree would test this.
  • Beyond the paper: a lightweight version that replaces the ODE solver with a closed-form temporal kernel could isolate how much of the improvement comes from continuous integration versus community-aware sampling.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes CTWalks, a representation-learning framework for continuous-time dynamic graphs that combines community-guided temporal walk sampling, community-enriched anonymization, and an ODE-based continuous encoding. The authors argue that the community-aware sampling removes locality bias, that the anonymization improves inductive generalization, and that the ODE encoder preserves irregular temporal dynamics. Experiments on five datasets compare CTWalks against DyRep, TGAT, TGN, CTDNE, JODIE, and CAWs in transductive and inductive temporal link prediction, with an ablation study and an additional static-graph experiment supporting the main components.

Significance. If the empirical claims are correct, CTWalks would provide a competitive, code-released baseline for CTDG link prediction, and the ablation study gives useful evidence that community-aware sampling and continuous integration matter. The theoretical sections are less convincing: Lemma 1 largely restates the transition restriction, and the matrix-factorization analysis in Appendix F does not match the actual sampling process. Most importantly, the dataset statistics in Table 1 and Appendix D.1 are mutually inconsistent for Taobao and Enron, so the headline state-of-the-art claim is not currently reproducible.

major comments (4)
  1. [Table 1 vs Appendix D.1] The dataset statistics are mutually inconsistent. Table 1 reports Taobao with 64,703 nodes and 77,436 temporal edges, while Appendix D.1 reports 987,994 nodes and 2,099,520 temporal edges; Enron is listed as 143 nodes/62,617 edges in Table 1 but 184 nodes/125,235 edges in Appendix D.1. Section 6.1 describes only chronological sorting, chronological splitting, and negative sampling, none of which can reconcile these numbers. Because Tables 2 and 7 report CTWalks as state of the art on these datasets, the reader cannot determine which graph was actually used. The authors must report the exact preprocessing pipeline, clarify which statistics are correct, and rerun or confirm all baselines on the same graph.
  2. [Section 5, Lemma 1 and Eq. (16)-(21)] The claim that CTWalks 'overcomes locality bias' is close to a restatement of the sampling restriction. The inequality in Lemma 1 is obtained by setting the first-visit probabilities of intra-community neighbors to zero and renormalizing by |Ninter(u)|; this is precisely the pruning built into Algorithm 1 for bridging nodes, not an independent theoretical property of the walk. The comparison is also against a uniform random walk on the full graph, whereas a more meaningful baseline would be a walk with the same restricted transition support. Moreover, Eq. (18) assumes that intra-community transitions can never reach nodes in other communities, which fails if the community partition is imperfect; the authors acknowledge this fragility in Section 7. Please reframe the lemma as a property of the construction or provide a comparison that does not assume the conclusion.
  3. [Appendix F, Eq. (48)-(56)] The matrix-factorization argument does not correspond to the sampling process in Algorithm 1. In CTWalks, each walk is generated entirely within either GI or GCi depending on the root node, so no single walk mixes MC and MI. The average transition behavior is therefore a weighted mixture of separate powers of MC and MI, not the sum 1/T sum_r (M_C^r + M_I^r) used in Eq. (48). Additionally, MC + MI is not a transition matrix because its rows generally sum to more than 1. As written, Lemma 2 does not establish the claimed connection between CTWalks and matrix factorization.
  4. [Abstract, Contribution 2, and Appendix D.3] The term 'parameter-free' is overstated. The sampling mechanism still depends on walk length l and number of walks per node C/R, both of which are tuned per dataset in Appendix D.3, and the ODE solver step size is manually fixed. If the intended claim is only that no direction-bias hyperparameter is introduced beyond the community partition, the text should state this qualification explicitly.
minor comments (6)
  1. [Definition 3 vs Sec. 4.2] Definition 3 states that temporal walks have strictly decreasing timestamps, while the anonymized walk construction in Sec. 4.2 uses ascending timestamps t1 < t2 < ... < tl; the two conventions should be aligned.
  2. [Appendix C.2] The batching analysis refers to 'Section B.1' when describing the batch-processing optimization, but the actual derivation appears in Appendix C.2; the cross-reference should be corrected.
  3. [Tables 2 and 7] Several entries in Tables 2 and 7 are difficult to read because bold and dagger markers are not consistently explained, and decimal alignment varies; please clarify the highlighting convention and format the tables uniformly.
  4. [Appendix D.2] The baseline name 'CA Ws' contains a spacing typo; it should read 'CAWs'.
  5. [Eq. (7)] In Eq. (7), the denominator is written as a sum over valid neighbors but it is not stated whether t' is the timestamp of the edge (v,u) and whether multiple edges between the same pair are treated separately; please clarify the indexing.
  6. [Appendix E] The static-graph experiment uses word2vec with walk length 80 and 10 walks per node, but the relationship of these settings to the temporal CTWalks configuration is not discussed; a sentence explaining the choice would improve reproducibility.

Circularity Check

2 steps flagged · score 6.0 of 10

Theoretical claims are definitional restatements (Lemma 1) and a known SGNS-PMI result (Appendix F); the empirical SOTA claim is external but Table 1 vs Appendix D.1 conflicts undermine reproducibility.

  1. self definitional [Section 4.1 Sampling Strategy and Section 5 Lemma 1 (Eqs. 16-21)]
    "Non-Bridging Nodes: Nodes confined within a single community Ci. Their temporal walks are restricted to the neighbors within the intra-community subgraph GCi. Bridging Nodes: Nodes that connect multiple communities. Their temporal walks are restricted to the neighbors within the inter-community subgraph GI, ensuring transitions only occur between bridging nodes. In CTWalk, intra-community transitions cannot reach nodes in different communities. Thus, for any j in Nintra(u), r^{t-1}_{jv}=0. Since |Ninter(u)| <= d(u), it follows that ..."

    Lemma 1 is presented as a theoretical proof that CTWalks 'reduces locality bias', but the inequality is installed by the sampling definition: bridging nodes are forbidden from intra-community neighbors (Nvalid(v)=N(v)∩GI), so the transition kernel is normalized by |Ninter(u)| and the claimed boost over a degree-normalized random walk is a restatement of that restriction. No independent mechanism is analyzed; the lemma would fail only if the sampling restriction were removed, which is exactly the design choice it is supposed to justify. The 'theoretical benefit' is therefore equivalent to the definition of the walk rather than a derived consequence.

  2. renaming known result [Appendix F, 'Analysis 2: Matrix Factorization Perspective of CTWalks', Lemma 2 and Eqs. (44)-(56)]
    "The Skip-Gram with negative sampling (SGNS) framework [57] in network embedding methods such as DeepWalk [51] and node2vec [50] has been shown to implicitly factorize a pointwise mutual information (PMI) matrix [58, 59]. CTWalks extends this perspective by incorporating a hierarchical, two-layer random walk that explicitly encodes both intra-community and inter-community transitions, resulting in a novel matrix factorization form."

    Once MC and MI are defined as the CTWalks inter-community and intra-community transition matrices, the 'correspondence' is the standard SGNS-PMI equivalence proved in [58,59] instantiated on that transition operator. The derivation relies on the walk being Skip-Gram based, so the factorization follows by definition of the PMI matrix, not from any CTWalks-specific structural property. Calling this a 'novel matrix factorization form' renames the known SGNS-PMI result with a composite transition matrix; it does not provide independent evidence for the method's superiority. This is a presentational renaming rather than a circular prediction, but it inflates the theoretical contribution.

full rationale

The paper's main empirical deliverable—AUC/AP improvements over baselines in Tables 2 and 7—is not circular: those numbers depend on data, baseline implementations, and training, and the method's equations do not force them. However, the two theory contributions advertised in the abstract are not independent. Lemma 1 defines bridging nodes as nodes whose walks are confined to GI and then proves that such walks are more likely to reach cross-community nodes; the inequality is the definition of the restriction, so the 'overcoming locality bias' result is a restatement. Appendix F derives the PMI factorization for MC+MI; since SGNS embeddings are already known to factor PMI [58,59], and MC/MI are the algorithm's own transition matrices, the claimed 'novel matrix factorization form' is the standard SGNS-PMI theorem restated, not an emergent property. Neither of these invalidates the empirical comparison, but they mean the paper's theoretical support is largely built in rather than derived. Separately, not a circularity issue but a reproducibility one: Table 1 reports Taobao with 64,703 nodes and 77,436 temporal edges while Appendix D.1 reports 987,994 nodes and 2,099,520 temporal edges, and Enron similarly differs (143/62,617 vs 184/125,235); this must be reconciled before the SOTA claim is verifiable. The 'parameter-free' claim is also undercut by the grid search over walk length and number of walks in Appendix D.3.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central empirical claim rests mainly on the assumption that communities detected on the aggregated weighted graph are stable and meaningful for future links, and on the fixed exponential time-decay kernel. The theoretical sections rely on additional ad hoc representations: time-interval rescaling and the M_C+M_I transition matrix in the factorization analysis. These are the unstated premises that a reader must accept before the claims follow.

free parameters (3)
  • Walk length l = tuned via grid search over {1,2,3}
    Appendix D.3 lists l as a tunable hyperparameter despite the 'parameter-free' claim; it controls context size of the anonymized walks.
  • Number of walks per node C = tuned via grid search over {16,32,64}
    Appendix D.3 lists C as tunable; it affects sampling variance and runtime.
  • ODE solver step size = 0.125 (fixed-step Runge-Kutta 3/8)
    Appendix D.3 sets this value by hand; it controls numerical accuracy and cost of the continuous integration.
assumptions (4)
  • domain assumption Weighted temporal graph aggregation preserves community structure relevant to future links.
    Eq. 6 aggregates a CTDG into a static weighted graph by counting interaction frequencies; the method assumes Louvain on this aggregate yields communities that remain valid for link prediction, including for nodes unseen at training time (Section 4.1, Section 6, Eq. 22).
  • ad hoc to paper Exponential time-decay with fixed unit rate is an adequate transition kernel.
    Eq. 7 sets the transition probability proportional to e^{-(t-t')} with no learned or tuned decay rate; this choice is not compared to alternatives.
  • ad hoc to paper Walk-interval reparameterization and logarithmic scaling preserve temporal dynamics.
    Appendix C.2 rescales each walk's time interval to [0,1] and applies log10(delta t + 1) to long intervals; the paper does not show the rescaled ODE yields the same representations as integrating on the original time axis.
  • ad hoc to paper CTWalks transition process is represented as M_C + M_I in the matrix-factorization analysis.
    Appendix F Eq. (48) uses a sum of intra- and inter-community transition matrices, whereas Algorithm 1 selects one subgraph per starting node; the factorized matrix is not the one actually sampled.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Community-Aware Temporal Walks: Parameter-Free Representation Learning on Continuous-Time Dynamic Graphs." pith.science (2026). https://pith.science/paper/6YER3VEW

@misc{pith2026250111880,
  author       = {Pith},
  title        = {Pith review of: Community-Aware Temporal Walks: Parameter-Free Representation Learning on Continuous-Time Dynamic Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6YER3VEW}},
  note         = {Machine review of arXiv:2501.11880}
}
read the original abstract

Dynamic graph representation learning plays a crucial role in understanding evolving behaviors. However, existing methods often struggle with flexibility, adaptability, and the preservation of temporal and structural dynamics. To address these issues, we propose Community-aware Temporal Walks (CTWalks), a novel framework for representation learning on continuous-time dynamic graphs. CTWalks integrates three key components: a community-based parameter-free temporal walk sampling mechanism, an anonymization strategy enriched with community labels, and an encoding process that leverages continuous temporal dynamics modeled via ordinary differential equations (ODEs). This design enables precise modeling of both intra- and inter-community interactions, offering a fine-grained representation of evolving temporal patterns in continuous-time dynamic graphs. CTWalks theoretically overcomes locality bias in walks and establishes its connection to matrix factorization. Experiments on benchmark datasets demonstrate that CTWalks outperforms established methods in temporal link prediction tasks, achieving higher accuracy while maintaining robustness.

Figures

Figures reproduced from arXiv: 2501.11880 by the authors.

Figure 1
Figure 1. Traditional anonymous encodings fail to differentiate [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the encoding mechanism in CTWalks. Each [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the data preparation process for link predic [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparison of temporal dynamics modeling approaches. Standard RNN maintains constant hidden states; RNN-Decay introduces [PITH_FULL_IMAGE:figures/full_fig_p022_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 68 canonical work pages

  1. [1]

    Temporal graph networks for deep learning on dynamic graphs,

    E. Rossi, B. Chamberlain, F. Frasca, D. Eynard, F. Monti, and M. Bronstein, “Temporal graph networks for deep learning on dynamic graphs,” in ICML 2020 Workshop on Graph Representation Learning, 2020

    work page 2020

  2. [2]

    Inductive representation learning on temporal graphs,

    D. Xu, C. Ruan, E. K¨orpeoglu, S. Kumar, and K. Achan, “Inductive representation learning on temporal graphs,” in 8th International Conference on Learning Represen- tations, OpenReview.net, 2020

    work page 2020

  3. [3]

    Inductive representation learning in temporal networks via causal anonymous walks,

    Y . Wang, Y .-Y . Chang, Y . Liu, J. Leskovec, and P. Li, “Inductive representation learning in temporal networks via causal anonymous walks,” in9th International Con- ference on Learning Representations, OpenReview.net, 2021

    work page 2021

  4. [4]

    Prov- ably expressive temporal graph networks,

    A. H. Souza, D. Mesquita, S. Kaski, and V . Garg, “Prov- ably expressive temporal graph networks,” in Advances in Neural Information Processing Systems (NeurIPS) , 2022

    work page 2022

  5. [5]

    A review of relational machine learning for knowl- edge graphs,

    M. Nickel, K. Murphy, V . Tresp, and E. Gabrilovich, “A review of relational machine learning for knowl- edge graphs,” Proceedings of the IEEE, vol. 104, no. 1, pp. 11–33, 2015

    work page 2015

  6. [6]

    Kernel of CycleGAN as a Principle homogeneous space

    S. Zhang, Z. Yang, M. Zhao, and G. Li, “Gnnrec: Graph neural network-based recommendation,” arXiv preprint arXiv:2001.09061, 2020

    work page Pith review arXiv 2001

  7. [7]

    Kgat: Knowledge graph attention network for recommenda- tion,

    X. Wang, X. He, Y . Cao, M. Liu, and T.-S. Chua, “Kgat: Knowledge graph attention network for recommenda- tion,” Proceedings of the 25th ACM SIGKDD Inter- national Conference on Knowledge Discovery & Data Mining, pp. 950–958, 2019

    work page 2019

  8. [8]

    Graph rep- resentation learning for biological networks,

    M. Zhang, W. Hu, J. Wang, and Z. Zhang, “Graph rep- resentation learning for biological networks,” in ACM SIGKDD International Conference on Knowledge Dis- covery & Data Mining, 2021

    work page 2021

Show all 73 references
  1. [9]

    Pro- tein interface prediction using graph convolutional net- works,

    A. Fout, J. Byrd, B. Shariat, and A. Ben-Hur, “Pro- tein interface prediction using graph convolutional net- works,” Advances in neural information processing sys- tems, vol. 30, pp. 6533–6542, 2017

    2017

  2. [10]

    Neighborhood-aware scalable tem- poral network representation learning,

    Y . Luo and P. Li, “Neighborhood-aware scalable tem- poral network representation learning,” in Learning on Graphs Conference, 2022

    2022

  3. [11]

    Towards better dy- namic graph learning: New architecture and unified li- brary,

    L. Yu, L. Sun, B. Du, and W. Lv, “Towards better dy- namic graph learning: New architecture and unified li- brary,” in Advances in Neural Information Processing Systems, 2023

    2023

  4. [12]

    Representation learning for dynamic graphs: A survey,

    S. M. Kazemi, R. Goel, K. Jain, I. Kobyzev, A. Sethi, P. Forsyth, and P. Poupart, “Representation learning for dynamic graphs: A survey,” Journal of Machine Learn- ing Research, vol. 21, no. 70, pp. 1–73, 2020

    2020

  5. [13]

    Dynamic graph representation learning with neural networks: A sur- vey,

    L. Yang, S. Adam, and C. Chatelain, “Dynamic graph representation learning with neural networks: A sur- vey,”arXiv preprint arXiv:2304.05729, 2023

    2023 arXiv

  6. [14]

    Encoder- decoder architecture for supervised dynamic graph learning: A survey,

    Y . Zhu, F. Lyu, C. Hu, X. Chen, and X. Liu, “Encoder- decoder architecture for supervised dynamic graph learning: A survey,” arXiv preprint arXiv:2203.10480, 2022

    2022 arXiv

  7. [15]

    Motif- aware representation learning on continuous-time dy- namic graphs,

    X. Jin, Q. Li, J. Zhou, W. Fan, and J. Tang, “Motif- aware representation learning on continuous-time dy- namic graphs,” arXiv preprint arXiv:2301.06434, 2023

    2023 arXiv

  8. [16]

    Motif-preserving dynamic attributed network embed- ding,

    M. Xia, S. Zhang, P. Zhou, F. Nie, and J. Huang, “Motif-preserving dynamic attributed network embed- ding,” IEEE Transactions on Knowledge and Data En- gineering, vol. 34, no. 12, pp. 5750–5763, 2021

    2021

  9. [17]

    Community detection in graphs,

    S. Fortunato, “Community detection in graphs,”Physics Reports, vol. 486, no. 3-5, pp. 75–174, 2010

    2010

  10. [18]

    Finding and evaluating community structure in networks,

    M. E. Newman and M. Girvan, “Finding and evaluating community structure in networks,” Physical review E , vol. 69, no. 2, p. 026113, 2004

    2004

  11. [19]

    A multiobjective evolution- ary algorithm based on similarity for community detec- tion from signed social networks,

    C. Liu, J. Liu, and Z. Jiang, “A multiobjective evolution- ary algorithm based on similarity for community detec- tion from signed social networks,” IEEE Transactions on Cybernetics, vol. 44, no. 12, pp. 2274 – 2287, 2014

    2014

  12. [20]

    Overlapping community detection in directed and undirected attributed networks using a multiobjective evolutionary algorithm,

    X. Teng, J. Liu, and M. Li, “Overlapping community detection in directed and undirected attributed networks using a multiobjective evolutionary algorithm,” IEEE Trans. Cybernetics , vol. 51, pp. 138 – 150, January 2021

    2021

  13. [21]

    Neural ordinary differential equations,

    R. T. Q. Chen, Y . Rubanova, J. Bettencourt, and D. Du- venaud, “Neural ordinary differential equations,” in Advances in Neural Information Processing Systems , vol. 31, 2018

    2018

  14. [22]

    Multivariate time series forecasting with dynamic graph neural odes,

    M. Jin, Y . Zheng, Y .-F. Li, S. Chen, B. Yang, and S. Pan, “Multivariate time series forecasting with dynamic graph neural odes,” arXiv preprint arXiv:2202.08408 , 2022

    2022 arXiv

  15. [23]

    Neu- ral ordinary differential equation control of dynamics on graphs,

    T. Asikis, L. B ¨ottcher, and N. Antulov-Fantulin, “Neu- ral ordinary differential equation control of dynamics on graphs,” arXiv preprint arXiv:2006.09773, 2020

    2006 arXiv

  16. [24]

    Signed graph neural ordinary differential equation for modeling continuous- time dynamics,

    L. Chen, K. Wu, J. Lou, and J. Liu, “Signed graph neural ordinary differential equation for modeling continuous- time dynamics,” arXiv preprint arXiv:2312.11198 , 2023

    2023 arXiv

  17. [25]

    Streaming dynamic graph neural net- works for continuous-time temporal graphs,

    Anonymous, “Streaming dynamic graph neural net- works for continuous-time temporal graphs,” IEEE Transactions on Neural Networks and Learning Sys- tems, 2022

    2022

  18. [26]

    Higher-order knowledge transfer for dynamic community detection with great changes,

    H. Ma, K. Wu, H. Wang, and J. Liu, “Higher-order knowledge transfer for dynamic community detection with great changes,” IEEE Transactions on Evolution- ary Computation, vol. 28, no. 1, pp. 90 – 104, 2024

    2024

  19. [27]

    Dynamic graph representa- tion learning via graph transformer networks,

    W. Cong, Y . Wu, Y . Tian, M. Gu, Y . Xia, M. Mah- davi, and C.-c. J. Chen, “Dynamic graph representa- tion learning via graph transformer networks,” CoRR, vol. abs/2111.10447, 2021

    2021 arXiv

  20. [28]

    dyngraph2vec: Capturing network dynamics using dynamic graph representation learning,

    P. Goyal, S. R. Chhetri, and A. Canedo, “dyngraph2vec: Capturing network dynamics using dynamic graph representation learning,” Knowledge-Based Systems , vol. 187, p. 104816, 2020

    2020

  21. [29]

    Evolvegcn: Evolving graph convolutional networks for dynamic graphs,

    A. Pareja, G. Domeniconi, J. Chen, T. Ma, T. Suzu- mura, H. Kanezashi, T. Kaler, T. B. Schardl, and C. E. Leiserson, “Evolvegcn: Evolving graph convolutional networks for dynamic graphs,” in Proceedings of the Thirty-Fourth AAAI Conference on Artificial Intelli- gence, pp. 536...

    2020

  22. [30]

    Dysat: Deep neural representation learning on dy- namic graphs via self-attention networks,

    A. Sankar, Y . Wu, L. Gou, W. Zhang, and H. Yang, “Dysat: Deep neural representation learning on dy- namic graphs via self-attention networks,” in Proceed- ings of the Thirteenth ACM International Conference on Web Search and Data Mining , pp. 519–527, ACM, 2020

    2020

  23. [31]

    Roland: graph learning framework for dynamic graphs,

    J. You, T. Du, and J. Leskovec, “Roland: graph learning framework for dynamic graphs,” in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discov- ery and Data Mining, pp. 2358–2366, ACM, 2022

    2022

  24. [32]

    Latent ordinary differential equations for irregularly-sampled time series,

    Y . Rubanova, R. T. Q. Chen, and D. Duvenaud, “Latent ordinary differential equations for irregularly-sampled time series,” in Advances in Neural Information Pro- cessing Systems, vol. 32, 2019

    2019

  25. [33]

    Continuous graph neural networks,

    L.-P. Xhonneux, M. Qu, and J. Tang, “Continuous graph neural networks,” Advances in Neural Information Pro- cessing Systems, vol. 33, pp. 18508–18519, 2020

    2020

  26. [34]

    The graph neural network model,

    F. Scarselli, M. Gori, A. C. Tsoi, M. Hagenbuchner, and G. Monfardini, “The graph neural network model,” IEEE Transactions on Neural Networks, vol. 20, no. 1, pp. 61–80, 2008

    2008

  27. [35]

    Autosgnn: Automatic propagation mechanism discovery for spec- tral graph neural networks,

    S. Mo, K. Wu, Q. Gao, X. Teng, and J. Liu, “Autosgnn: Automatic propagation mechanism discovery for spec- tral graph neural networks,” in AAAI, 2025

    2025

  28. [36]

    Finding structure in time,

    J. L. Elman, “Finding structure in time,” Cognitive Sci- ence, vol. 14, no. 2, pp. 179–211, 1990

    1990

  29. [37]

    Predicting dy- namic embedding trajectory in temporal interaction net- works,

    S. Kumar, X. Zhang, and J. Leskovec, “Predicting dy- namic embedding trajectory in temporal interaction net- works,” inProceedings of the 25th ACM SIGKDD Inter- national Conference on Knowledge Discovery & Data Mining, pp. 1269–1278, ACM, 2019

    2019

  30. [38]

    Fast algorithm for detecting commu- nity structure in networks,

    M. E. Newman, “Fast algorithm for detecting commu- nity structure in networks,” Physical review E, vol. 69, no. 6, p. 066133, 2004

    2004

  31. [39]

    Fast unfolding of communities in large networks,

    V . D. Blondel, J.-L. Guillaume, R. Lambiotte, and E. Lefebvre, “Fast unfolding of communities in large networks,” Journal of Statistical Mechanics: Theory and Experiment, vol. 2008, no. 10, p. P10008, 2008

    2008

  32. [40]

    Community detection in networks with node attributes,

    J. Yang, J. McAuley, and J. Leskovec, “Community detection in networks with node attributes,” ICDM, pp. 1151–1156, 2013

    2013

  33. [41]

    Overlapping community de- tection at scale: a nonnegative matrix factorization ap- proach,

    J. Yang and J. Leskovec, “Overlapping community de- tection at scale: a nonnegative matrix factorization ap- proach,” Proceedings of the sixth ACM international conference on Web search and data mining , pp. 587– 596, 2013

    2013

  34. [42]

    Community detection in attributed graphs: An embedding approach,

    B. Du, Z. Wu, and J. Pei, “Community detection in attributed graphs: An embedding approach,” Proceed- ings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, pp. 589–598, 2017

    2017

  35. [43]

    Dyrep: Learning representations over dynamic graphs,

    R. Trivedi, M. Farajtabar, P. Biswal, and H. Zha, “Dyrep: Learning representations over dynamic graphs,” in International Conference on Learning Rep- resentations (ICLR), 2019

    2019

  36. [44]

    Continuous-time dynamic network embed- dings,

    G. Nguyen, J. Lee, R. Rossi, N. Ahmed, E. Koh, and S. Kim, “Continuous-time dynamic network embed- dings,” Proceedings of the ACM International Con- ference on Web Search and Data Mining (WSDM) , pp. 373–381, 2018

    2018

  37. [45]

    Snap datasets: Stanford large network dataset collection

    J. Leskovec and A. Krevl, “Snap datasets: Stanford large network dataset collection.” http://snap.stanford. edu/data, 2014. Accessed: 2016-01-01

    2014

  38. [46]

    Learning tree-based deep model for recommender sys- tems,

    H. Zhu, X. Li, P. Zhang, G. Li, J. He, H. Li, and K. Gai, “Learning tree-based deep model for recommender sys- tems,” in Proceedings of the 24th ACM SIGKDD Inter- national Conference on Knowledge Discovery & Data Mining, pp. 1079–1088, ACM, 2018

    2018

  39. [47]

    Learning phrase representations using rnn encoder-decoder for statistical machine translation,

    K. Cho, B. Van Merrienboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, and Y . Bengio, “Learning phrase representations using rnn encoder-decoder for statistical machine translation,” in Proceedings of the 2014 Conference on Empirical Methods in Natural Lan- guage Proc...

    2014

  40. [48]

    Neural spatio- temporal point processes,

    R. T. Chen, B. Amos, and M. Nickel, “Neural spatio- temporal point processes,” in International Conference on Learning Representations, 2020

    2020

  41. [49]

    Ef- ficient estimation of word representations in vector space,

    T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Ef- ficient estimation of word representations in vector space,” in Proceedings of the International Conference on Learning Representations (ICLR), 2013

    2013

  42. [50]

    node2vec: Scalable fea- ture learning for networks,

    A. Grover and J. Leskovec, “node2vec: Scalable fea- ture learning for networks,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 855–864, ACM, 2016

    2016

  43. [51]

    DeepWalk: On- line learning of social representations,

    B. Perozzi, R. Al-Rfou, and S. Skiena, “DeepWalk: On- line learning of social representations,” in Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pp. 701–710, ACM, 2014

    2014

  44. [52]

    LINE: Large-scale information network embedding,

    J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan, and Q. Mei, “LINE: Large-scale information network embedding,” in Proceedings of the 24th International Conference on World Wide Web, pp. 1067–1077, 2015

    2015

  45. [53]

    struc2vec: Learning node representations from struc- tural identity,

    L. F. Ribeiro, P. H. Saverese, and D. R. Figueiredo, “struc2vec: Learning node representations from struc- tural identity,” in Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Dis- covery and Data Mining, pp. 385–394, ACM, 2017

    2017

  46. [54]

    Inductive rep- resentation learning on large graphs,

    W. Hamilton, Z. Ying, and J. Leskovec, “Inductive rep- resentation learning on large graphs,” in Proceedings of Advances in Neural Information Processing Systems, pp. 1024–1034, 2017

    2017

  47. [55]

    Commu- nity preserving network embedding,

    X. Wang, P. Cui, J. Wang, J. Pei, and W. Zhu, “Commu- nity preserving network embedding,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 31, pp. 203–209, 2017

    2017

  48. [56]

    The network data repos- itory with interactive graph analytics and visualization,

    R. A. Rossi and N. K. Ahmed, “The network data repos- itory with interactive graph analytics and visualization,”

  49. [57]

    Distributed representations of words and phrases and their compositionality,

    T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, “Distributed representations of words and phrases and their compositionality,” in Advances in Neural Information Processing Systems (NeurIPS) , pp. 3111–3119, 2013

    2013

  50. [58]

    Neural word embedding as implicit matrix factorization,

    O. Levy and Y . Goldberg, “Neural word embedding as implicit matrix factorization,” in Proceedings of Advances in Neural Information Processing Systems , pp. 2177–2185, 2014

    2014

  51. [59]

    Network embedding as matrix factorization: Unify- ing DeepWalk, LINE, PTE, and node2vec,

    J. Qiu, Y . Dong, H. Ma, J. Li, K. Wang, and J. Tang, “Network embedding as matrix factorization: Unify- ing DeepWalk, LINE, PTE, and node2vec,” inProceed- ings of the 11th ACM International Conference on Web Search and Data Mining, pp. 459–467, 2018

    2018

  52. [60]

    A critical review of recurrent neural networks for sequence learn- ing,

    Z. C. Lipton, J. Berkowitz, and C. Elkan, “A critical review of recurrent neural networks for sequence learn- ing,” arXiv preprint arXiv:1506.00019, 2015

    2015 arXiv

  53. [61]

    Recurrent neural networks for multivariate time series with missing values,

    Z. Che, S. Purushotham, K. Cho, D. Sontag, and Y . Liu, “Recurrent neural networks for multivariate time series with missing values,” inScientific reports, vol. 8, pp. 1– 12, Nature Publishing Group, 2018

    2018

  54. [62]

    Neural hawkes process: A neurally self-modulating multivariate point process,

    H. Mei and J. Eisner, “Neural hawkes process: A neurally self-modulating multivariate point process,” in Advances in Neural Information Processing Systems , pp. 6754–6764, 2017. A Notations and Definitions Symbol Definition G = {(ei, ti)}N i=1 A continuous-time dynamic graph (...

    2017

  55. [64]

    Local Optimization Phase: Initially, each node is treated as a separate community. The algorithm iteratively moves nodes between neighboring communities to maximize the modularity gain ∆Q, defined as: ∆Q = P in +2kin 2m − P tot +ktot 2m 2 − P tot 2m 2 , (23) where P in is the ...

  56. [65]

    The process is repeated until the modularity Q no longer improves significantly

    Community Aggregation Phase: Once no further modularity improvement is possible, each community is treated as a supernode, forming a new graph. The process is repeated until the modularity Q no longer improves significantly. The Louvain method is highly scalable and effective,...

  57. [66]

    For each node wi, the hidden state hi is computed as: hi = g(h′ i−1, A(wi)), (24) where: • h′ i−1 is the cumulative hidden state from the previous step

    Instantaneous Update Function g The function g handles the discrete, step-wise state updates at specific nodes along a temporal walk. For each node wi, the hidden state hi is computed as: hi = g(h′ i−1, A(wi)), (24) where: • h′ i−1 is the cumulative hidden state from the previ...

  58. [67]

    Unlikeg, which incorporates node-specific input features,f focuses solely on temporal dynamics and acts on the output ofg

    Continuous Temporal Evolution Function f The function f models the continuous evolution of the hidden state over time intervals between nodes. Unlikeg, which incorporates node-specific input features,f focuses solely on temporal dynamics and acts on the output ofg. The cumulat...

  59. [68]

    All temporal edges are sorted chronologically to maintain temporal consistency

  60. [69]

    Edges are split into training (70%), validation (15%), and testing (15%) sets

  61. [70]

    Negative sampling is performed to generate edges absent in the original graph, ensuring a balanced dataset. D.2 Baselines and Hyperparameter Tuning Baseline Methods: To evaluate the performance of our proposed method, we compare it with six state-of-the-art baselines specifica...

  62. [71]

    New-Old: Interactions between unseen and observed nodes

  63. [72]

    New-New: Interactions exclusively between unseen nodes. Hardware Configuration: Experiments are conducted on an Ubuntu Linux server equipped with four NVIDIA GeForce RTX 3090 GPUs, two Intel(R) Xeon(R) Silver 4210 CPUs (2.20GHz), and 252GB of RAM. Each experiment is repeated f...

  64. [73]

    Anonymized Walk Construction

    Store the final anonymized representation A(w; Su, Sv, Cu, Cv) for all w ∈ Vinteraction. Anonymized Walk Construction. After anonymizing individual nodes, the anonymized temporal walk is constructed. For a single walk W = {w1, w2, . . . , wl} with timestamps t1 < t2 < · · ·< t...

  65. [2015]

    Available at https://networkrepository.com

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.