Pith. sign in

REVIEW 4 major objections 4 minor 40 references

Never Skip a Batch: Dense Learning of Temporal GNNs via Adaptive Pseudo-Supervision

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read History-averaged pseudo-labels reduce gradient variance in temporal GNN training, cutting time-to-accuracy by up to 15 times on standard benchmarks.

desk verdict A plausible pseudo-labeling trick for temporal GNNs, but the variance proof is flawed and the headline speedups rely on a nonstandard 5% training split. read the letter →

arxiv 2505.12526 v2 pith:VKJ2W63R submitted 2025-05-18 cs.LG

Alexander Panyshev , Dmitry Vinichenko , Oleg Travkin , Roman Alferov , Alexey Zaytsev This is my paper · ORCID
classification cs.LG
keywords temporalgraphnetworkspseudo-labelinglabelsparsitygradientvariancereductionconvergenceaccelerationdynamiclearningmovingaveragelabelsBenchmark
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

In temporal graphs, most training batches arrive without labels, so models update memory but do not learn. This paper claims that such gaps can be filled with cheap pseudo-targets: for each node, average (or exponentially average, or copy) the labels observed in that node's past, and train on those targets as if they were real. The central theoretical assertion is that this aggregation lowers the variance of stochastic gradients, and that under slowly changing preferences the convergence speed improves by a factor of order $\min(h,k)$, where $h$ is how much history is aggregated and $k$ is the number of classes. Empirically, using the TGNv2 model on the four Temporal Graph Benchmark datasets, the paper reports reaching top validation performance up to 15 times faster while keeping accuracy competitive. The appeal is that the method adds no parameters, leaves the loss and architecture untouched, and turns previously idle batches into productive updates.

What carries the argument

The central object is the history-averaged pseudo-label $\tilde{y}_t^{(v)}$, an aggregate of all supervision signals observed for node $v$ before time $t$; the abstract calls this Moving-Averaged Labels (MAL), the body calls it History-Averaged Labels (HAL). Three variants are used: Historical Average (plain mean of all past targets), Moving Average (exponentially weighted recent labels), and Persistent Forecast (last observed label). The pseudo-label enters the objective by replacing $y$ in the cross-entropy loss while the model, loss, and architecture stay unchanged. The theoretical load is carried by Lemma 2, which models one component of the pseudo-label as the product of an observation-indicator Bernoulli $\xi$ and a scaled multinomial frequency $\eta_k$, derives the variance $\mathrm{Var}(t_h)=u(k-1)/(k^2 h)+u(1-u)/k^2$, and feeds it into the SGD regret bound of Theorem 1 to produce the $\min(h,k)$ speedup factor.

What would settle it

Train the method on a synthetic temporal graph where each node's true label is redrawn uniformly at random at every timestamp, so historical averages carry no information about the current label; if convergence is still faster than vanilla SGD, the reported speedup is not explained by unbiased variance reduction.

Watch

Extended reading notes

Core claim

The paper's discovery, stated in good faith, is that the slow drift of node preferences is itself a usable supervisory signal. When ground-truth targets are missing, replacing them with the node's historical label distribution (computed as a historical average, an exponential moving average, or the last observed value) gives an unbiased estimator of the current target, and the extra averaging shrinks the per-label variance from order $1/k$ to order $1/\min(h,k)$. Substituting this variance estimate into the standard SGD regret bound yields an upper bound on regret that is smaller by the same factor, which the paper reads as a proof that pseudo-labels accelerate convergence. On the TGB suite, TGNv2 with these pseudo-targets reaches its best validation NDCG@10 in fewer steps and less wall-clock time, up to 15 times faster on the harder datasets, with test quality on par with or slightly better than the vanilla model.

Load-bearing premise

The result depends on historical pseudo-labels being an unbiased proxy for a node's current label and on label variance being proportional to gradient variance; if past preferences stop tracking current ones, or if label noise does not translate linearly into gradient noise, the claimed convergence speedup does not follow.

Editorial extensions

If this is right

  • On label-sparse temporal graphs, training becomes continuous: batches without ground truth still produce gradient updates instead of being skipped.
  • Time-to-accuracy shrinks by up to an order of magnitude on the tested benchmarks, which matters when models must be retrained frequently on streaming data.
  • Because the method adds no parameters and changes neither the loss nor the architecture, it can be dropped into any temporal GNN, not just TGNv2.
  • The three aggregation variants provide a stability-versus-reactivity dial: persistent forecast suits stable preferences, moving average suits drifting ones, and historical average suits long-run baselines.
  • The theoretical factor $\min(h,k)$ predicts diminishing returns from extending history once history length exceeds the number of target categories.

Reading between the lines

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

  • Outside the paper's experiments, the same pseudo-supervision logic could apply to link prediction and temporal knowledge graph completion, where supervision is also sparse and sequential.
  • A testable extension is to make the aggregation window adaptive: choose per-node weighting from validation performance, which could handle both stable and volatile users without adding parameters.
  • The paper's own target-shuffling ablation points to a boundary: when preferences drift fast, persistent forecast degrades while moving average retains more quality, so a drift-aware mixture of the three variants is a natural next step.
  • If the variance argument transfers to other architectures, label-level variance reduction would complement optimizer-level methods such as Adam or variance-reduced SGD, attacking a different component of optimization noise.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes to train temporal graph networks on every batch by using pseudo-labels computed from past observed targets, with three aggregation strategies: Historical Average (HA), Moving Average (MA), and Persistent Forecast (PF). It claims a theoretical result that historical averaging reduces gradient variance and accelerates convergence by a factor of min(h,k), where h is the history length and k the number of classes, and it reports experiments on four Temporal Graph Benchmark datasets showing up to 15x faster training for TGNv2 with competitive accuracy. The method introduces no trainable parameters and is intended to be architecture-agnostic. A notable feature of the experimental protocol is that for three of the four datasets only the last 5% of the chronologically ordered training edges are retained, which the authors justify by saying that on the full training sets models converge within a single epoch.

Significance. If the claims were correct, the paper would offer a simple, zero-parameter, architecture-agnostic technique for mitigating sparse supervision in temporal GNNs, and the ablation studies on edge-order and target-order shuffling are useful. The appendix contains complete proofs of the stated lemmas and theorems, and the method description is mostly clear. However, the theoretical derivation rests on a label model that does not match the implemented averaging process and on an unproved proportionality between label variance and gradient variance; moreover, the headline speedups are measured on a modified benchmark with a nonstandard 5% training split and no comparison to standard TGB results. These issues undermine the central claims of the paper as stated, so the contribution is not established in its current form.

major comments (4)
  1. [Section 3.2, Lemma 2 (Appendix A, Lemma 4)] The random variable t_h = eta_k * xi, with eta_k independent of xi, does not model the historical average of one-hot labels under missingness. For h independent observations in which a fixed label is observed with probability u/k per observation, the component average is Binomial(h, u/k)/h, whose variance is u(k-u)/(k^2 h) and vanishes as h grows. The paper's formula u(k-1)/(k^2 h) + u(1-u)/k^2 retains an h-independent term u(1-u)/k^2, so the claimed min(h,k) speedup in Theorem 3 follows from a label model that the method does not implement. The theorem therefore does not establish the stated convergence acceleration; the variance-reduction result must be re-derived for the actual averaging process, and with the correct model the speedup factor would be h rather than min(h,k).
  2. [Section 3.2, paragraph after Lemma 2] The assertion that the gradient variance can be written as \tilde{c} Var(t_h) for a non-stochastic constant \tilde{c} across all layers is not justified. For the last layer, Var(p_i - y_i) is not equal to Var(y_i) in general; it also depends on the variance of the model output p_i through the joint distribution of inputs and labels. For earlier layers, the chain rule introduces products of activations and Jacobians that are random and correlated with the label, so a label-independent constant \tilde{c} is not available without additional assumptions. Since Theorem 3 uses this proportionality to convert the label variance into a gradient-variance bound, the convergence-rate claim is not supported by the provided derivation.
  3. [Section 5.1, Table 1; Section 5.2, Table 2] The reported speedups for tgbn-genre, tgbn-reddit, and tgbn-token are obtained on a modified protocol that keeps only the last 5% of the chronologically ordered training edges, not on the standard TGB 70/15/15 split. The paper explicitly states that on the full training sets models converge within one epoch, which would obscure the effect of pseudo-labeling. Consequently, the abstract's unqualified claims of up to 15x acceleration and 'new SOTA' are not established for the standard benchmark; the experiments compare methods only in an artificial low-data regime. No results on the full training splits, no comparison with published TGB numbers, and no error bars or multiple-seed statistics are reported, so even within the modified regime the quantitative claims are not verified.
  4. [Section 4.1, Eqs. (5)-(6); Section 4.2] The manuscript is internally inconsistent about the training target. The supervised-batch description says predictions are compared to ground-truth labels y_t^{(v)}, but Eq. (6) states that the loss replaces y_t^{(v)} with pseudo-labels \tilde{y}_t^{(v)} 'for almost all nodes and batches.' It is therefore unclear whether supervised batches use true labels, pseudo-labels, or a mixture; this determines whether the pseudo-label gradient is unbiased and is essential for reproducing the experiments. The method definition must be clarified before the empirical results can be interpreted.
minor comments (4)
  1. [Abstract and title] The metadata abstract refers to 'Moving-Averaged Labels (MAL)' and to experiments on TGNv2 and DyRepv2, while the full-text abstract refers to 'History-Averaged Labels (HAL)' and the experiments cover only TGNv2; these inconsistencies should be corrected.
  2. [Section 4.2] The paper says the method has zero additional parameters, but the noise scale gamma, the noise amplitude alpha, the moving-average window w, and the 5% truncation threshold are tunable hyperparameters; their selection procedure and sensitivity are not reported.
  3. [Section 5.2] The 'N steps' metric in Table 2 is not defined precisely; it should state whether it counts gradient updates, batches, or forward passes.
  4. [Figure 3] The logarithmic time axis makes the claimed speedup difficult to read; the figure would benefit from marking the convergence point for each method and from showing the Default curve beyond the early training range.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the theoretical result is derived from an explicit label model and the empirical claims are measured against baselines, not fitted.

full rationale

The paper's central theoretical claim is that aggregating historical one-hot labels into a running average reduces stochastic-gradient variance and thus improves the SGD regret bound. The derivation chain is self-contained: Lemma 2 defines a probabilistic label model (t_h = eta_k * xi, with eta_k a scaled Binomial component and xi a Bernoulli missingness indicator), computes its mean and variance in closed form, and Theorem 3 plugs that variance into the standard regret bound from Shamir and Zhang. No step reduces to its own input by construction: the variance reduction is not assumed, but computed; the speedup factor min(h,k) follows from comparing the upper bounds for h=1 and h>1; and the pseudo-label update rules in Section 4 (Historical Average, Moving Average, Persistent Forecast) are defined independently of the convergence claim. The empirical evaluation reports measured NDCG@10 and wall-clock time for HA/MA/PF against a Default baseline; no fitted parameter is renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work. The nonstandard 5% training split is an external-validity concern rather than a circularity, and the mismatch between the Lemma 2 label model and the true historical average is a correctness concern, not a circular reduction. The paper also contains no load-bearing self-citation. Accordingly, no circular step is identified.

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

The paper introduces no new physical or conceptual entities. The central empirical claims depend on the free parameters listed above, particularly the 5% truncation and unstated hyperparameters, while the theoretical claims rest on the axioms above, none of which are established beyond assertion.

free parameters (3)
  • 5% training truncation = 5% of chronologically ordered edges
    Section 5.1: the last 5% of training edges are retained; this threshold was empirically determined through ablation studies, changing the benchmark and making pseudo-labeling appear more effective.
  • Moving average window w = Not reported for main results
    Section 5.3 analyzes window sizes, but Table 2 does not state which w was used for the MA results on each dataset.
  • Noise scaling gamma = Not reported
    Introduced in Eq. (7) as a noise scaling factor, but never mentioned in the experimental setup or tables, leaving the actual value unspecified.
assumptions (4)
  • domain assumption SGD regret bound for strongly convex loss applies to TGN training with pseudo-labels
    Invoked in Section 3.1 and used without verifying convexity or unbiasedness of the pseudo-label gradient with respect to the true loss.
  • ad hoc to paper Gradient variance is proportional to label variance: Var(dL/dc_ij) = e_j^2 Var(t_h), and similarly for all layers with a constant c
    Assumed in Section 3.2 after computing Var(t_h); the extension to hidden layers is asserted 'due to the chain rule' without proof.
  • ad hoc to paper Historical average label is modeled as product of independent Bernoulli(u) and Binomial(h, 1/k)/h
    Lemma 2 in the appendix uses this model, which does not match the true average of one-hot labels with missingness; the independence and the retained variance term are not justified.
  • domain assumption User preferences are constant or change slowly during the label history
    Stated in Section 3.2 and used for the variance analysis; contradicted by the temporal drift motivation and by the target-shuffling ablation in Section 5.4, where shuffling targets severely degrades performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Never Skip a Batch: Dense Learning of Temporal GNNs via Adaptive Pseudo-Supervision." pith.science (2026). https://pith.science/paper/VKJ2W63R

@misc{pith2026250512526,
  author       = {Pith},
  title        = {Pith review of: Never Skip a Batch: Dense Learning of Temporal GNNs via Adaptive Pseudo-Supervision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VKJ2W63R}},
  note         = {Machine review of arXiv:2505.12526}
}
read the original abstract

Temporal graph networks suffer from irregular supervision in realworld dynamic graphs, as most minibatches contain few labeled events. The lack of labels leads to high-variance gradient updates and, consequently, slow wall-clock convergence. To constructively reduce sparsity, our Moving-Averaged Labels (MAL) assigns soft pseudo-targets based on past supervised signals using a running label distribution while leaving the loss and the model architecture unchanged. Thus, supervision gaps are replaced with informative signals independent of a temporal graph model and the message passing or memory components used. Theoretical analysis supports our insight that aggregating historical supervision into moving average targets reduces stochastic gradient variance, yielding faster convergence under mild assumptions. Experimentally, for TGNv2 and DyRepv2 (our modification of DyRep) models, MAL boosts predictive performance, establishing a new SOTA, and improves time-to-accuracy (on average 6x faster to reach the top score) for a common suite of Temporal Graph Benchmark datasets.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 29 canonical work pages

  1. [1]

    Arazo, D

    E. Arazo, D. Ortego, P. Albert, N. E. O’Connor, and K. McGuin- ness. Pseudo-labeling and confirmation bias in deep semi-supervised learning. In 2020 International joint conference on neural networks (IJCNN), pages 1–8. IEEE, 2020

    work page 2020

  2. [2]

    Bertin-Mahieux, D

    T. Bertin-Mahieux, D. P. Ellis, B. Whitman, and P. Lamere. The million song dataset. In Ismir, volume 2, page 10, 2011

    work page 2011

  3. [3]

    Bottou, F

    L. Bottou, F. E. Curtis, and J. Nocedal. Optimization methods for large- scale machine learning. SIAM Review, 60(2):223–311, 2018. doi: 10. 1137/16M1080173

    work page 2018

  4. [4]

    S. Bubeck. Convex optimization: Algorithms and complexity. Foun- dations and Trends® in Machine Learning, 8(3–4):231–357, 2015. doi: 10.1561/2200000050

    doi:10.1561/2200000050 2015

  5. [5]

    R. Cai, X. Bai, Z. Wang, Y . Shi, P. Sondhi, and H. Wang. Modeling sequential online interactive behaviors with temporal point process. In Proceedings of the 27th ACM international conference on information and knowledge management, pages 873–882, 2018

    work page 2018

  6. [6]

    W. Cong, S. Zhang, J. Kang, B. Yuan, H. Wu, X. Zhou, H. Tong, and M. Mahdavi. Do we really need complicated model architectures for temporal networks?, 2023

    work page 2023

  7. [7]

    S. Deng, H. Rangwala, and Y . Ning. Learning dynamic context graphs for predicting social events. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, pages 1007–1016, 2019

    work page 2019

  8. [8]

    Gastinger, S

    J. Gastinger, S. Huang, M. Galkin, E. Loghmani, A. Parviz, F. Pour- safaei, J. Danovitch, E. Rossi, I. Koutis, H. Stuckenschmidt, et al. TGB 2.0: A benchmark for learning on temporal knowledge graphs and het- erogeneous graphs. Advances in neural information processing systems, 37:140199–140229, 2024

    work page 2024

Show all 40 references
  1. [9]

    S. Guo, Y . Lin, N. Feng, C. Song, and H. Wan. Attention based spatial-temporal graph convolutional networks for traffic flow forecast- ing. In Proceedings of the AAAI conference on artificial intelligence , volume 33, pages 922–929, 2019

    2019

  2. [10]

    G. Han, W. Chen, X. Zhang, J. Xu, A. Liu, and L. Zhao. Tkgf-ntp: Temporal knowledge graph forecasting via neural temporal point pro- cess. In 2023 IEEE International Conference on Web Services (ICWS), pages 318–328. IEEE, 2023

    2023

  3. [11]

    Huang, F

    S. Huang, F. Poursafaei, J. Danovitch, M. Fey, W. Hu, E. Rossi, J. Leskovec, M. Bronstein, G. Rabusseau, and R. Rabbany. Temporal graph benchmark for machine learning on temporal graphs. Advances in Neural Information Processing Systems, 36:2056–2073, 2023

    2023

  4. [12]

    R. J. Hyndman and G. Athanasopoulos. Forecasting: principles and practice. OTexts, 2018

    2018

  5. [13]

    Järvelin and J

    K. Järvelin and J. Kekäläinen. Cumulated gain-based evaluation of ir techniques. ACM Transactions on Information Systems (TOIS), 20(4): 422–446, 2002

    2002

  6. [14]

    Johnson and T

    R. Johnson and T. Zhang. Accelerating stochastic gradient descent us- ing predictive variance reduction. In Advances in Neural Information Processing Systems, pages 315–323, 2013

    2013

  7. [15]

    D. P. Kingma and J. Ba. Adam: A method for stochastic optimization. International Conference on Learning Representations (ICLR), 2015

    2015

  8. [16]

    Klenitskiy, A

    A. Klenitskiy, A. V olodkevich, A. Pembek, and A. Vasilev. Does it look sequential? an analysis of datasets for evaluation of sequential recom- mendations. In Proceedings of the 18th ACM Conference on Recom- mender Systems, pages 1067–1072, 2024

    2024

  9. [17]

    Kumar, X

    S. Kumar, X. Zhang, and J. Leskovec. Predicting dynamic embed- ding trajectory in temporal interaction networks. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, pages 1269–1278, 2019

    2019

  10. [18]

    Lee et al

    D.-H. Lee et al. Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks. In Workshop on challenges in representation learning, ICML, volume 3, page 896. Atlanta, 2013

    2013

  11. [19]

    Y . Li, J. Yin, and L. Chen. Informative pseudo-labeling for graph neural networks with few labels. Data Mining and Knowledge Discovery , 37 (1):228–254, 2023

    2023

  12. [20]

    Z. Ma, K. Ethayarajh, T. Thrush, S. Jain, L. Wu, R. Jia, C. Potts, A. Williams, and D. Kiela. Dynaboard: An evaluation-as-a-service plat- form for holistic next-generation benchmarking. Advances in Neural Information Processing Systems, 34:10351–10367, 2021

    2021

  13. [21]

    Pfaff, M

    T. Pfaff, M. Fortunato, A. Sanchez-Gonzalez, and P. Battaglia. Learning mesh-based simulation with graph networks. In International confer- ence on learning representations, 2020

    2020

  14. [22]

    Rakhlin, O

    A. Rakhlin, O. Shamir, and K. Sridharan. Making gradient descent op- timal for strongly convex stochastic optimization. InProceedings of the 29th International Coference on International Conference on Machine Learning, pages 1571–1578, 2012

    2012

  15. [23]

    Robbins and S

    H. Robbins and S. Monro. A stochastic approximation method. The Annals of Mathematical Statistics, 22(3):400–407, 1951. doi: 10.1214/ aoms/1177729586

    1951

  16. [24]

    Rossi, B

    E. Rossi, B. Chamberlain, F. Frasca, D. Eynard, F. Monti, and M. Bron- stein. Temporal graph networks for deep learning on dynamic graphs. arXiv preprint arXiv:2006.10637, 2020

    2006 arXiv

  17. [25]

    Sanchez-Gonzalez, J

    A. Sanchez-Gonzalez, J. Godwin, T. Pfaff, R. Ying, J. Leskovec, and P. Battaglia. Learning to simulate complex physics with graph net- works. In International conference on machine learning, pages 8459–

  18. [26]

    Shalev-Shwartz and S

    S. Shalev-Shwartz and S. Ben-David. Understanding machine learning: From theory to algorithms. Cambridge university press, 2014

    2014

  19. [27]

    Shamir and T

    O. Shamir and T. Zhang. Stochastic gradient descent for non-smooth optimization: Convergence results and optimal averaging schemes. In Proceedings of the 30th International Conference on Machine Learn- ing, ICML 2013, Atlanta, GA, USA, 16-21 June 2013 , volume 28 of JMLR Works...

    2013

  20. [28]

    Shamsi, F

    K. Shamsi, F. Victor, M. Kantarcioglu, Y . Gel, and C. G. Akcora. Char- talist: Labeled graph datasets for utxo and account-based blockchains. Advances in Neural Information Processing Systems, 35:34926–34939, 2022

    2022

  21. [29]

    W. Song, Z. Xiao, Y . Wang, L. Charlin, M. Zhang, and J. Tang. Session- based social recommendation via dynamic graph attention networks. In Proceedings of the Twelfth ACM international conference on web search and data mining, pages 555–563, 2019

    2019

  22. [30]

    B. A. Tjandra, F. Barbero, and M. Bronstein. Enhancing the expressivity of temporal graph networks through source-target identification. arXiv preprint arXiv:2411.03596, 2024

    2024 arXiv

  23. [31]

    Trivedi, M

    R. Trivedi, M. Farajtabar, P. Biswal, and H. Zha. Dyrep: Learning repre- sentations over dynamic graphs. In International Conference on Learn- ing Representations, 2019

    2019

  24. [32]

    Wang and J

    H. Wang and J. Leskovec. Unifying graph convolutional neural net- works and label propagation. arXiv preprint arXiv:2002.06755, 2020

    2002 arXiv

  25. [33]

    Wang, Y .-Y

    Y . Wang, Y .-Y . Chang, Y . Liu, J. Leskovec, and P. Li. Inductive repre- sentation learning in temporal networks via causal anonymous walks. arXiv preprint arXiv:2101.05974, 2021

    2021 arXiv

  26. [34]

    Xie, M.-T

    Q. Xie, M.-T. Luong, E. Hovy, and Q. V . Le. Self-training with noisy student improves imagenet classification. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10687–10698, 2020

    2020

  27. [35]

    D. Xu, C. Ruan, E. Korpeoglu, S. Kumar, and K. Achan. In- ductive representation learning on temporal graphs. arXiv preprint arXiv:2002.07962, 2020

    2002 arXiv

  28. [36]

    Y . You, Z. Zhang, C.-J. Hsieh, J. Demmel, and K. Keutzer. ImageNet training in minutes. In Proceedings of the 47th international conference on parallel processing, pages 1–10, 2018

    2018

  29. [37]

    L. Yu, L. Sun, B. Du, and W. Lv. Towards better dynamic graph learn- ing: New architecture and unified library. In Thirty-seventh Conference on Neural Information Processing Systems, 2023

    2023

  30. [38]

    Zhan and C

    K. Zhan and C. Niu. Mutual teaching for graph convolutional networks. Future Generation Computer Systems, 115:837–843, 2021

    2021

  31. [39]

    L. Zhao, Y . Song, C. Zhang, Y . Liu, P. Wang, T. Lin, M. Deng, and H. Li. T-GCN: A temporal graph convolutional network for traffic prediction. IEEE transactions on intelligent transportation systems , 21(9):3848– 3858, 2019

    2019

  32. [40]

    X. Zhu. Semi-supervised learning with graphs. Carnegie Mellon Uni- versity, 2005. A Proofs of the theoretical results For convenience, we repeat below the statements we aim to prove. Lemma 4. The expectation of th is u k and the variance of th isuk−1 k2h +u(1−u) 1 k2 . Proof. ...

    2005

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.