Pith. sign in

REVIEW 4 major objections 5 minor 42 references

Temporal Neighbourhood Aggregation: Predicting Future Links in Temporal Graphs via Recurrent Variational Graph Convolutions

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A temporal graph model, Temporal Neighbourhood Aggregation, predicts new edges in future graph snapshots up to 23% more accurately than competing deep baselines while using fewer parameters.

desk verdict A genuinely new architecture with strong real-data results, but the evaluation write-up has documentation gaps that need fixing before the headline numbers can be trusted. read the letter →

arxiv 1908.08402 v2 pith:3GCU4WDS submitted 2019-08-21 cs.SI

classification cs.SI
keywords temporalgraphrepresentationlearningdynamiclinkpredictionconvolutionalnetworksrecurrentneuralvariationalinferencenext-graphneighbourhoodaggregation
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

The paper introduces Temporal Neighbourhood Aggregation (TNA), a vertex-embedding model that reads a series of graph snapshots and tries to predict the next graph in the series. Each TNA block combines a graph-convolution layer with a gated recurrent unit, and stacking two blocks lets a vertex draw on temporal change in both its one-hop and two-hop neighbourhoods. No vertex features, labels, or random walks are required. The final embeddings are sampled from a Gaussian using variational inference, and the next adjacency matrix is decoded by a parameter-free inner product. On three real-world temporal networks the model predicts previously unseen edges more accurately than the compared deep baselines, with the paper reporting gains of up to 23% on real-world datasets while using fewer parameters than the recurrent baseline family.

What carries the argument

The central object is the TNA block: one graph-convolution layer followed by a gated recurrent unit, with a concatenated skip connection and layer normalisation. In the block, the GCN aggregates a vertex's current one-hop neighbourhood, the GRU reads the resulting vertex representations across snapshots so the block carries memory of how that neighbourhood has been changing, and the linear mixing layer learns how much weight to give topology versus time at each depth. Stacking two TNA blocks aggregates temporal information from two hops, and the variational sampling head, a pair of GCN layers parameterising mean and variance with the reparameterisation trick, produces the final embedding. The generative decoder is simply the sigmoid inner product of embedding pairs, which is why the model has no parameterised decoder and only 133K–239K parameters on the test graphs.

What would settle it

Run TNA on the Wiki-Vote or UCI-Messages snapshots exactly as Algorithm 1 is written, without adding a vertex-alignment step; if the fixed $|V|\times|V|$ inputs and the new-edge evaluation are undefined or the reported AUC/AP cannot be reproduced, the claim that the model needs no preprocessing is false.

Watch

Extended reading notes

Core claim

TNA's central claim is that a temporally recurrent graph convolution can capture how each vertex's neighbourhood changes at several depths and convert that directly into an accurate forecast of the next graph. The model computes, for each snapshot, a graph-convolved representation, passes it through a GRU that carries state across snapshots, and mixes the two with a learned linear combination after layer normalisation; two such blocks stacked in depth give each vertex access to two hops of temporal neighbourhood change. Variational sampling then produces the embedding matrix $Z_t$ from two GCN heads that estimate mean and variance, and the future graph is reconstructed as a sigmoid inner product of $Z_t$ with itself, so the decoder has no trainable parameters. The training objective is the variational lower bound on $\log p(A_{t+1}|Z_t)$ minus the KL divergence to a Gaussian prior, plus a small L2 penalty. In the paper's experiments TNA attains the highest mean AUC and average precision for new-edge prediction on Bitcoin-Alpha, Wiki-Vote, and UCI Messages at every tested fraction of the time history, with 133K–239K parameters, while the DynAE/DynRNN/DynAERNN baselines use millions.

Load-bearing premise

The pipeline assumes the vertex set is fixed and identically ordered across all snapshots, so a single $|V|\times|V|$ adjacency matrix and identity feature matrix can represent every graph.

Editorial extensions

If this is right

  • Because the decoder is parameter-free, total model size is set by the encoder alone, which is what allows the method to scale to graphs of thousands of vertices and long snapshot histories without the memory cost of recurrent decoders.
  • The ablation shows that including second-hop temporal information raises AUC and AP by roughly 10 percentage points over first-hop-only on Bitcoin-Alpha, so multi-depth recurrence is the component that carries most of the predictive gain.
  • Since no features, labels, or random walks are needed, the model applies directly to any snapshot graph sequence, not just graphs with rich node attributes.
  • The multi-step forecasting experiment indicates that feeding the model's own predicted graph back in as input keeps new-edge prediction accuracy above that of the closest baseline for several steps ahead.
  • On full-graph reconstruction, TNA remains competitive on Bitcoin-Alpha and Wiki-Vote and slightly behind on UCI, showing the gain on new edges does not come from ignoring existing edges.

Reading between the lines

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

  • As an extension, we would expect the same architecture to need an explicit vertex-alignment or inductive aggregation step on graphs where vertices appear and disappear, because the fixed $|V|\times|V|$ input and identity features are only well-defined for a stable vertex set.
  • We would also test how sensitive the 23% gain is to snapshot granularity; monthly batching of Bitcoin-Alpha versus weekly batching of UCI changes how much edge evolution each recurrent step sees, and the model's advantage over D-GCN may shrink when snapshots are so fine that few new edges appear per step.
  • A natural component study to run elsewhere is the full ablation from the paper's Table IV on Wiki-Vote and UCI; if variational sampling or the skip connection contributes less on those graphs, the one-architecture-fits-all claim would need qualification.
  • A testable variant would swap the fixed-graph convolution for an aggregator that handles previously unseen vertices; that would separate the contribution of temporal recurrence from the contribution of spectral convolution.
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 / 5 minor

Summary. The paper proposes Temporal Neighbourhood Aggregation (TNA), a temporal graph embedding architecture that stacks GCN layers with GRU recurrence at each hop, adds layer normalization and skip connections, and produces vertex representations via variational sampling. The model is trained end-to-end to predict the next temporal snapshot using an inner-product decoder with zero learnable parameters. The authors evaluate TNA on three real-world datasets (Bitcoin-Alpha, Wiki-Vote, UCI-Messages) and two synthetic datasets, comparing it against static and temporal baselines (GAE, GVAE, TO-GAE, TO-GVAE, DynAE, DynRNN, DynAERNN, D-GCN). The central claim is that TNA outperforms competing methods at predicting new temporal edges by up to 23% on real datasets while using fewer model parameters. The paper also includes ablation studies showing the contribution of each architectural component.

Significance. If the reported results hold under a clearly specified and reproducible evaluation protocol, TNA would be a useful parameter-efficient baseline for temporal link prediction, with the interesting design choice of placing recurrence at each neighborhood depth rather than only at the graph level. The paper ships a public code repository, and the ablation study in Table IV is a useful contribution. However, the significance of the empirical claims is currently limited by an underspecified evaluation protocol, and the model's ability to handle time-varying vertex sets is not established. The core new-edge prediction claim therefore cannot yet be assessed as reliable.

major comments (4)
  1. [Section III-C, Table I, Equations (1) and (4); Section VI] The model as specified assumes a fixed, consistently ordered vertex set. Table I sets X to the identity matrix of A, so the first GCN weight matrix in Eq. (1) has input dimension |V|, and Eq. (3) applies the same shared-parameter TNA block to every snapshot. However, Eq. (4) indexes the variational distribution by |V_t|, and the datasets Wiki and UCI contain vertices that appear after the initial snapshots. Section VI explicitly lists 'enabling vertex arrival to be modelled' as future work, confirming that the current model does not handle arrival. Algorithm 1 contains no preprocessing step that constructs a common vertex universe, trims appearing vertices, or otherwise reconciles time-varying vertex sets with shared weights. The authors must specify exactly how this was handled; otherwise the reported AUC/AP numbers are not reproducible, and options such as padding with zero rows could leak information about future vertex existence into earlier snapshots.
  2. [Section IV-A, Algorithm 1, Section V-B] The evaluation protocol for new-edge prediction is underspecified with respect to negative sampling. The task is defined as predicting edges in E_t \ E_{t-1}, and AUC and AP are computed, but the paper never states how negative examples are sampled for these metrics. This is load-bearing because AUC and AP are ranking measures over a positive set and a negative set; different negative-sampling strategies (e.g., random non-edges, same-degree non-edges, or non-edges with geographic constraints) can substantially change the reported scores and the comparison with baselines. The authors should describe the negative sampling distribution, the ratio of negatives to positives, and whether the same negative set is used for all methods.
  3. [Table III, Section V-B] The meaning of the '25%, 50%, 100%' columns in Table III is ambiguous and the text does not define it. Algorithm 1 describes iterating over all G_t with t ≥ 3 and training on G_1,...,G_{t-1}, but Table III reports results 'when predicting at various percentages of the length of the time-sequence.' It is unclear whether the models are trained on only the first 25%/50% of snapshots and then evaluated on all later snapshots, or whether each test graph is predicted using only the immediately preceding fraction of history. These different protocols lead to different difficulties and different numbers. The authors should define the training/test split precisely and relate it to Algorithm 1.
  4. [Section V-B, Table III] The paper states that TNA 'significantly outperforms' the baselines, but no statistical significance tests are reported. The results are presented as means with standard deviations over (apparently) three random seeds, and for several cells the standard deviations overlap substantially (e.g., Wiki AUC at 100%: TNA 0.634 ± 0.050 vs D-GCN 0.563 ± 0.087, where the overlap is not huge but exists; Bitcoina AUC at 25%: TNA 0.665 ± 0.067 vs D-GCN 0.622 ± 0.084). At minimum, the authors should state the number of runs and report paired significance tests (or confidence intervals) for the headline comparisons, and they should avoid the word 'significant' unless such tests support it.
minor comments (5)
  1. [Throughout] There are numerous typos and slight formatting issues, including 'scientific', 'varitonal', 'Iternational', 'disregard' (in the conclusion), and inconsistent capitalization of 'Variational' in Table IV; these should be corrected.
  2. [Section IV-D] The D-GCN baseline is described only briefly as 'a dynamic GCN similar to approaches in [20] and [21].' Since D-GCN is a primary competitor in Table III and Figure 4, the authors should provide more implementation detail (depth, hidden sizes, training schedule) or refer to a public implementation.
  3. [Table V] For the SBM dataset, TNA's AUC and AP are reported as 0.502 ± 0.024 for both metrics, which is suspiciously identical and likely a typo; the same applies to the R-Cora row. The authors should check these entries.
  4. [Section III-E] The paragraph describing the training procedure mentions 'full-batch gradient descent with RMSProp,' but RMSProp is a stochastic optimizer; 'full-batch' is ambiguous and should be clarified (e.g., whether each update uses all vertices in the snapshot).
  5. [Section IV-C, Table II] The dataset table reports 'First Edge' and 'Last Edge' dates, but the text says Wiki-Vote covers '28th March 2004' while the table says '28/02/2005'; this inconsistency should be fixed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: TNA predicts held-out future snapshots from past snapshots; self-citations are not load-bearing.

full rationale

The derivation chain is not circular. TNA's predictive step is defined by Eq. (3)-(6): the variational distribution q(Z_t|X_t,A_t) is conditioned on the current snapshot, and the generative model p(A_{t+1}|Z_t) = prod sigma_s(z_i z_j^T) uses the future adjacency matrix only as the reconstruction target inside the ELBO, never as conditioning input. Algorithm 1 trains a fresh model on G1..G_{t-1} and evaluates AUC/AP on the held-out new edges E_t\E_{t-1} of G_t. The reported 23% margin is therefore an empirical comparison on data not used to fit the model, not a fitted parameter renamed as a prediction. Self-citations are present but not load-bearing: TO-GAE/TO-GVAE from the authors' prior work [34] are only baselines, and [38]-[39] are used for synthetic graph rewiring; neither supplies the theoretical premise that TNA predicts future edges. No uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via citation. The main caveat is the fixed-vertex-set assumption: Table I sets X to the identity matrix of A, all TNA parameters are shared through time, and Eq. (4) indexes |V_t|, so if Wiki/UCI contain vertex arrivals, the shared GCN weight dimension is undefined without a preprocessing step that Algorithm 1 does not describe. The conclusion itself concedes that 'enabling vertex arrival to be modelled' is future work. This is a reproducibility/correctness limitation, not circularity, because it does not reduce the prediction target to the model's input by construction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard neural network machinery and a set of dataset assumptions; no new physical entities are introduced. The main burden is the fixed-vertex-set assumption and the choice of hyperparameters.

free parameters (5)
  • Number of stacked TNA blocks = 2
    Selected via grid search; controls the neighbourhood depth from which temporal information is aggregated (Section III-E, V-A).
  • Hidden dimensions of TNA blocks = 32 (first block), 16 (second block)
    Selected via grid search; sets model capacity and parameter count (Section III-E).
  • Learning rate = 0.001
    Chosen empirically as best for RMSProp full-batch training (Section III-E).
  • L2 regularization coefficient (lambda) = 1e-5
    Set in the regularizer of Eq. 7 to reduce overfitting (Section III-D).
  • Training epochs = 200
    Empirically found to give best results (Section III-E).
assumptions (5)
  • standard math GCN propagation rule (Eq. 1) and GRU update equations (Eq. 2) from prior literature.
    The model is built directly on these standard formulations; no proof is given because they are taken from Kipf & Welling and Cho et al.
  • standard math Variational lower bound (Eq. 6) with a Gaussian prior is a valid objective for the inference model.
    Uses the VAE ELBO as in Kingma & Welling; the KL term is tractable for Gaussian priors.
  • domain assumption Each snapshot G_t has a fixed, consistently ordered vertex set V with symmetric adjacency matrix.
    The model uses full-batch updates on |V|x|V| matrices for every snapshot; no handling of dynamic vertex sets is described (Section III-A, III-C).
  • domain assumption Identity matrix as input features (X=I) encodes vertex identity without external attributes.
    The paper sets X to the identity and relies on the network to distinguish vertices purely by connectivity and position (Table I, Section III-C).
  • domain assumption The inner-product decoder (Eq. 5) is a sufficient generative model for the next adjacency matrix.
    The decoder uses zero learnable parameters; the premise is that embedding inner products after a sigmoid capture edge probabilities (Section III-C).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Temporal Neighbourhood Aggregation: Predicting Future Links in Temporal Graphs via Recurrent Variational Graph Convolutions." pith.science (2026). https://pith.science/paper/3GCU4WDS

@misc{pith2026190808402,
  author       = {Pith},
  title        = {Pith review of: Temporal Neighbourhood Aggregation: Predicting Future Links in Temporal Graphs via Recurrent Variational Graph Convolutions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3GCU4WDS}},
  note         = {Machine review of arXiv:1908.08402}
}
read the original abstract

Graphs have become a crucial way to represent large, complex and often temporal datasets across a wide range of scientific disciplines. However, when graphs are used as input to machine learning models, this rich temporal information is frequently disregarded during the learning process, resulting in suboptimal performance on certain temporal infernce tasks. To combat this, we introduce Temporal Neighbourhood Aggregation (TNA), a novel vertex representation model architecture designed to capture both topological and temporal information to directly predict future graph states. Our model exploits hierarchical recurrence at different depths within the graph to enable exploration of changes in temporal neighbourhoods, whilst requiring no additional features or labels to be present. The final vertex representations are created using variational sampling and are optimised to directly predict the next graph in the sequence. Our claims are reinforced by extensive experimental evaluation on both real and synthetic benchmark datasets, where our approach demonstrates superior performance compared to competing methods, out-performing them at predicting new temporal edges by as much as 23% on real-world datasets, whilst also requiring fewer overall model parameters.

Figures

Figures reproduced from arXiv: 1908.08402 by the authors.

Figure 1
Figure 1. The temporal link prediction task is to predict the new [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of the Temporal Neighbourhood Aggre [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The overall Temporal Neighbourhood Aggregation Model: two stacked TNA blocks learning both topological and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: AUC and AP for the Wiki and UCI datasets when predicting new edges [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 28 canonical work pages

  1. [1]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in International Conference on Learning Rep- resentations (ICLR), 2017

  2. [2]

    Egonet: identification of human disease ego-network modules,

    R. Yang, Y . Bai, Z. Qin, and T. Yu, “Egonet: identification of human disease ego-network modules,” BMC genomics , vol. 15, no. 1, p. 314, 2014

  3. [3]

    Moleculenet: a benchmark for molecular machine learning,

    Z. Wu, B. Ramsundar, E. N. Feinberg, J. Gomes, C. Geniesse, A. S. Pappu, K. Leswing, and V . Pande, “Moleculenet: a benchmark for molecular machine learning,” Chemical science, vol. 9, no. 2, pp. 513– 530, 2018

  4. [4]

    Interaction networks for learning about objects, relations and physics,

    P. Battaglia, R. Pascanu, M. Lai, D. J. Rezende et al. , “Interaction networks for learning about objects, relations and physics,” in Advances in Neural Information Processing Systems , 2016, pp. 4502–4510

  5. [5]

    node2vec : scalable feature learning for networks,

    A. Grover and J. Leskovec, “node2vec : scalable feature learning for networks,” International Conference on Knowledge Discovery and Data Mining, 2016

  6. [6]

    Graph Embedding Techniques, Applications, and Performance: A Survey

    P. Goyal and E. Ferrara, “Graph embedding techniques, applications, and performance: a survey,” arXiv preprint arXiv:1705.02801 , 2017

  7. [7]

    Pytorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga et al. , “Pytorch: An imperative style, high-performance deep learning library,” in Advances in Neural Information Processing Systems , 2019, pp. 8024–8035

  8. [8]

    Laplacian eigenmaps and spectral techniques for embedding and clustering,

    M. Belkin and P. Niyogi, “Laplacian eigenmaps and spectral techniques for embedding and clustering,” Advances in neural information process- ing systems, pp. 585–591, 2002

Show all 42 references
  1. [9]

    Distributed large-scale natural graph factorization,

    A. Ahmed, N. Shervashidze, S. Narayanamurthy, V . Josifovski, and A. J. Smola, “Distributed large-scale natural graph factorization,” In- ternational conference on World Wide Web , pp. 37–48, 2013

  2. [10]

    DeepWalk: online learning of social representations,

    B. Perozzi, R. Al-Rfou, and S. Skiena, “DeepWalk: online learning of social representations,” International Conference on Knowledge Discov- ery and Data Mining , 2014

  3. [11]

    Autoencoders, unsupervised learning, and deep architectures,

    P. Baldi, “Autoencoders, unsupervised learning, and deep architectures,” in Proceedings of ICML workshop on Unsupervised and Transfer Learning, 2012, pp. 37–49

  4. [12]

    Variational graph auto-encoders,

    T. N. Kipf and M. Welling, “Variational graph auto-encoders,” arXiv preprint arXiv:1611.07308, 2016

  5. [13]

    Stwalk: learning trajectory representations in temporal graphs,

    S. Pandhre, H. Mittal, M. Gupta, and V . N. Balasubramanian, “Stwalk: learning trajectory representations in temporal graphs,” in Proceedings of the ACM India Joint International Conference on Data Science and Management of Data . ACM, 2018, pp. 210–219

  6. [14]

    Netwalk: A flexible deep embedding approach for anomaly detection in dynamic networks,

    W. Yu, W. Cheng, C. C. Aggarwal, K. Zhang, H. Chen, and W. Wang, “Netwalk: A flexible deep embedding approach for anomaly detection in dynamic networks,” in Iternational Conference on Knowledge Discovery & Data Mining . ACM, 2018, pp. 2672–2681

  7. [15]

    Continuous-time dynamic network embeddings,

    G. H. Nguyen, J. B. Lee, R. A. Rossi, N. K. Ahmed, E. Koh, and S. Kim, “Continuous-time dynamic network embeddings,” in 3rd International Workshop on Learning Representations for Big Networks (WWW BigNet), 2018

  8. [16]

    Dyngem: Deep embedding method for dynamic graphs,

    P. Goyal, N. Kamra, X. He, and Y . Liu, “Dyngem: Deep embedding method for dynamic graphs,” arXiv preprint arXiv:1805.11273 , 2018

  9. [17]

    Structural deep network embedding,

    D. Wang, P. Cui, and W. Zhu, “Structural deep network embedding,” in International Conference on Knowledge Discovery and Data Mining . ACM, 2016, pp. 1225–1234

  10. [18]

    Net2net: Accelerating learning via knowledge transfer,

    T. Chen, I. Goodfellow, and J. Shlens, “Net2net: Accelerating learning via knowledge transfer,” arXiv preprint arXiv:1511.05641 , 2015

  11. [19]

    dyngraph2vec: Captur- ing network dynamics using dynamic graph representation learning,

    P. Goyal, S. R. Chhetri, and A. Canedo, “dyngraph2vec: Captur- ing network dynamics using dynamic graph representation learning,” Knowledge-Based Systems, 2019

  12. [20]

    Dynamic graph convolutional networks,

    F. Manessi, A. Rozza, and M. Manzo, “Dynamic graph convolutional networks,” Pattern Recognition, p. 107000, 2019

  13. [21]

    Structured sequence modeling with graph convolutional recurrent networks,

    Y . Seo, M. Defferrard, P. Vandergheynst, and X. Bresson, “Structured sequence modeling with graph convolutional recurrent networks,” in International Conference on Neural Information Processing . Springer, 2018, pp. 362–373

  14. [22]

    Gcn-gan: A non-linear temporal link prediction model for weighted dynamic networks,

    K. Lei, M. Qin, B. Bai, G. Zhang, and M. Yang, “Gcn-gan: A non-linear temporal link prediction model for weighted dynamic networks,” in IEEE INFOCOM 2019-IEEE Conference on Computer Communications. IEEE, 2019, pp. 388–396

  15. [23]

    Gc-lstm: Graph convo- lution embedded lstm for dynamic link prediction,

    J. Chen, X. Xu, Y . Wu, and H. Zheng, “Gc-lstm: Graph convo- lution embedded lstm for dynamic link prediction,” arXiv preprint arXiv:1812.04206, 2018

  16. [24]

    Evolvegcn: Evolving graph convolutional networks for dynamic graphs,

    A. Pareja, G. Domeniconi, J. Chen, T. Ma, T. Suzumura, H. Kanezashi, T. Kaler, and C. E. Leisersen, “Evolvegcn: Evolving graph convolutional networks for dynamic graphs,” arXiv preprint arXiv:1902.10191 , 2019

  17. [25]

    Revisiting spatial- temporal similarity: A deep learning framework for traffic prediction,

    H. Yao, X. Tang, H. Wei, G. Zheng, and Z. Li, “Revisiting spatial- temporal similarity: A deep learning framework for traffic prediction,” in AAAI Conference on Artificial Intelligence , 2019

  18. [26]

    Diffusion convolutional re- current neural network: Data-driven traffic forecasting,

    Y . Li, R. Yu, C. Shahabi, and Y . Liu, “Diffusion convolutional re- current neural network: Data-driven traffic forecasting,” arXiv preprint arXiv:1707.01926, 2017

  19. [27]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

  20. [28]

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

    K. Cho, B. Van Merri ¨enboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, and Y . Bengio, “Learning phrase representations using rnn encoder-decoder for statistical machine translation,” arXiv preprint arXiv:1406.1078, 2014

  21. [29]

    Fastgcn: fast learning with graph convolutional networks via importance sampling,

    J. Chen, T. Ma, and C. Xiao, “Fastgcn: fast learning with graph convolutional networks via importance sampling,” arXiv preprint arXiv:1801.10247, 2018

  22. [30]

    Inductive representation learning on large graphs,

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

  23. [31]

    Layer normalization,

    J. L. Ba, J. R. Kiros, and G. E. Hinton, “Layer normalization,” arXiv preprint arXiv:1607.06450, 2016

  24. [32]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

  25. [33]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114, 2013

  26. [34]

    Temporal graph offset reconstruction: Towards tempo- rally robust graph representation learning,

    S. Bonner, J. Brennan, I. Kureshi, G. Theodoropoulos, A. S. McGough, and B. Obara, “Temporal graph offset reconstruction: Towards tempo- rally robust graph representation learning,” in 2018 IEEE International Conference on Big Data (Big Data) . IEEE, 2018, pp. 3737–3746

  27. [35]

    SNAP Datasets: Stanford large network dataset collection,

    J. Leskovec and A. Krevl, “SNAP Datasets: Stanford large network dataset collection,” http://snap.stanford.edu/data, Jun. 2014

  28. [36]

    Konect: the koblenz network collection,

    J. Kunegis, “Konect: the koblenz network collection,” in Proceedings of the 22nd International Conference on World Wide Web . ACM, 2013, pp. 1343–1350

  29. [37]

    Stochastic blockmodels and community structure in networks,

    B. Karrer and M. E. Newman, “Stochastic blockmodels and community structure in networks,” Physical review E , vol. 83, no. 1, p. 016107, 2011

  30. [38]

    Deep topology classification: A new approach for massive graph classification,

    S. Bonner, J. Brennan, G. Theodoropoulos, I. Kureshi, and A. S. McGough, “Deep topology classification: A new approach for massive graph classification,” in International Conference on Big Data . IEEE, 2016, pp. 3290–3297

  31. [39]

    Efficient comparison of massive graphs through the use of graph fingerprints,

    S. Bonner, J. Brennan, I. Kureshi, M. Stephen, and G. Theodoropoulos, “Efficient comparison of massive graphs through the use of graph fingerprints,” in KDD Workshop on Mining and Learning with Graphs (MLG), 2016

  32. [40]

    Dynamicgem: A library for dynamic graph embedding methods,

    P. Goyal, S. R. Chhetri, N. Mehrabi, E. Ferrara, and A. Canedo, “Dynamicgem: A library for dynamic graph embedding methods,” arXiv preprint arXiv:1811.10734, 2018

  33. [41]

    E-lstm-d: A deep learning framework for dynamic network link pre- diction,

    J. Chen, J. Zhang, X. Xu, C. Fu, D. Zhang, Q. Zhang, and Q. Xuan, “E-lstm-d: A deep learning framework for dynamic network link pre- diction,” arXiv preprint arXiv:1902.08329 , 2019

  34. [42]

    Generating text with recurrent neural networks,

    I. Sutskever, J. Martens, and G. E. Hinton, “Generating text with recurrent neural networks,” in Proceedings of the 28th International Conference on Machine Learning (ICML-11) , 2011, pp. 1017–1024

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.