Pith. sign in

REVIEW 6 major objections 5 minor 52 references

STAHGNet: Modeling Hybrid-grained Heterogenous Dependency Efficiently for Traffic Prediction

T0 review · 6 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A recurrent graph-attention network claims the best reported accuracy on four traffic datasets while using at least four times less GPU memory than prior state-of-the-art graph models.

desk verdict A plausible efficiency-first traffic forecasting architecture whose SOTA claims are not yet verifiable because the paper under-specifies the neighbor sampling and botches two equations. read the letter →

arxiv 2412.17524 v1 pith:ZHIF3JYG submitted 2024-12-23 cs.AI

Jiyao Wang , Zehua Peng , Yijia Zhang , Dengbo He , Lei Chen This is my paper · ORCID
classification cs.AI
keywords trafficflowpredictiongraphattentionnetworkhybrid-granularitymodelingspatio-temporaldependencyrandomneighborsamplingmultivariatetimeseriesefficientneuralPeMSdatasets
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 tries to establish that traffic flow prediction improves when a model captures both fine-grained, moment-to-moment dependencies between nearby road sensors and coarse-grained, global temporal patterns, in a single recurrent architecture. It proposes STAHGNet, which couples these two granularities through a Hybrid Graph Attention Module and a Coarse-grained Temporal Graph generator, and claims that this coupling beats twelve baseline and state-of-the-art methods on four public PeMS datasets. The paper also argues that random sampling of K neighbor nodes, instead of processing the full road graph, preserves the dependencies needed for accuracy while cutting GPU memory from several thousand megabytes to about one gigabyte. If correct, this would make accurate spatio-temporal traffic forecasting substantially cheaper to train and deploy.

What carries the argument

The central object is the STAHGNet recurrent cell, which at every timestamp updates a target node's representation using information from K randomly sampled neighboring sensors. Inside the cell, a Hybrid Graph Attention Module (HGAT) computes a static spatial aggregation from an inverse-distance adjacency matrix masked to the sampled neighbors, combines it with fine-grained temporal attention scores between the target and each neighbor, and then propagates updated representations back to the neighbors. A separate Coarse-grained Temporal Graph (CTG) generator computes an attenuation-weighted sum of all timestep representations, builds a sparse implicit adjacency matrix through a softmax over pairwise similarities, and applies a final graph convolution to fuse coarse-grained global information with the fine-grained recurrent output. The random sampling step is what reduces input storage from $O(B \times N \times w)$ to $O(B \times K \times w)$, which is the mechanism behind the paper's memory-efficiency claim.

What would settle it

Retrain STAHGNet on PeMSD4 with K=4 but force the sampled neighbor set to include the fifth-nearest sensor instead of a nearer one; if prediction error for that target changes substantially, the model depends on which neighbors are sampled rather than on the graph structure, and if it does not change, the random-sampling assumption is validated.

Watch

Extended reading notes

Core claim

The paper claims that hybrid-granularity spatio-temporal dependency—combining continuous fine-grained temporal evolution at each timestamp with static spatial adjacency and coarse-grained global temporal structure—is what lets a traffic forecaster outperform models that use only one granularity. STAHGNet encodes each target sensor and its K sampled neighbors in separate recurrent cells, aggregates neighbor information with a heterogenous attention mechanism that fuses inverse-distance spatial weights and per-timestamp attention scores, and then builds an implicit coarse-grained temporal graph using an attenuation-weighted sum of all timestep representations. On PeMSD3, PeMSD4, PeMSD7, and PeMSD8, the reported results show the lowest MAE, RMSE, and MAPE among the compared methods (e.g., MAE 14.82 on PeMSD3 and 18.92 on PeMSD4), and the reported GPU memory is 1087 MB versus 5423 MB for the best prior graph model, a fourfold or larger saving. Ablations show that removing the static spatial matrix, the coarse temporal graph, or the sampled neighbor information degrades performance, which the paper reads as evidence that each component is load-bearing.

Load-bearing premise

The load-bearing premise is that keeping only K randomly sampled neighbors, and fabricating extra neighbors for low-degree nodes, preserves every spatial-temporal dependency that actually drives traffic, so the model never needs the missing links.

Editorial extensions

If this is right

  • If random K-neighbor sampling preserves the dependencies that matter, full-graph GNNs are not necessary for traffic forecasting, and memory use scales with the sampled neighborhood size instead of the number of sensors.
  • The reported ablations imply that both static spatial adjacency and fine-grained temporal attention are needed; removing either one lowers accuracy on all four datasets.
  • The multi-step prediction results suggest the hybrid-granularity model maintains its advantage over SOTA baselines for most of the 12-step horizon, with the gap narrowing only at later steps.
  • The learned evolving dependency matrices could provide an interpretable view of which sensor pairs become correlated during congestion events, beyond what a static road graph shows.
  • The grid-search results indicate that longer historical windows help on longer datasets and that more sampled hops help when the data has a higher missing ratio.

Reading between the lines

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

  • The paper's efficiency claim suggests that the effective dependency graph of traffic is sparse, and similar random-neighbor sampling could be transplanted to other large-scale graph time-series tasks where full-graph message passing is the bottleneck.
  • The reported memory savings, if reproduced, would make the model feasible for deployment on edge devices or for city-scale networks with thousands of sensors, a direction the paper only hints at.
  • The CTG's learned sparse adjacency matrix could serve as a diagnostic tool for transportation agencies, highlighting non-obvious congestion-propagation links that are not captured by physical road distance.
  • Because the hyperparameters K, w, and H are chosen by grid search on validation data, an implicit untested assumption is that these choices transfer across time periods and missing-data patterns; a testable extension would be to re-tune them on temporally disjoint validation segments.
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

6 major / 5 minor

Summary. The paper proposes STAHGNet, a recurrent graph-based model for traffic flow prediction. The model combines a Hybrid Graph Attention Module (HGAT) that aggregates K randomly sampled neighboring road nodes per timestep with a Coarse-granularity Temporal Graph (CTG) generator that constructs an implicit similarity graph from attenuated sums of hidden states. Additional feature engineering adds normalized and change-rate series to the raw flow input. The authors claim that STAHGNet outperforms eight classical baselines and four SOTA methods on four PeMS datasets in MAE, RMSE, and MAPE, while reducing GPU memory to roughly one quarter of the best prior graph model. The paper also includes sensitivity tests, ablations, a multi-step prediction comparison, and a case study of computational cost.

Significance. If the results hold, the main contribution is an efficiency-accuracy trade-off: replacing full-graph message passing with K-neighbor sampling and small recurrent cells reportedly yields lower errors than full-graph GNN baselines while using 1087 MB versus 5423 MB for MS-GAT. The ablations cover the main components, and the multi-step experiments are a useful addition. However, the central empirical claim is compromised by internal contradictions, incomplete mathematical definitions, and an underspecified sampling procedure. Because the accuracy and memory results both depend on the sampling and compensation mechanism, and because the paper does not provide enough detail to reconstruct the model, the current manuscript does not support its headline claims.

major comments (6)
  1. [§3.2.3, Eq. (7)] The attenuation coefficient γ(t)=1/(w−t) is undefined at t=w if t ranges over the window positions 1,...,w. As written, the last timestep in each window produces a division by zero. Please redefine the coefficient or the index range, and clarify the intended summation limits.
  2. [§3.2.3, Eq. (8)] Equation (8) is syntactically incomplete: the expression At = σ((log(θij/(1−θij)) is missing a closing parenthesis and does not contain the softmax operation that the surrounding text describes. The text states that a non-parameterized softmax function produces a sparse adjacency matrix, but the displayed formula is a logistic function without row normalization or sparsification. Please provide the exact computation, including how θij is derived from E and how sparsity is enforced.
  3. [§3.2.1] The random neighbor sampling procedure is not specified sufficiently for reproduction. The text says that K neighbor nodes are reserved and that low-degree nodes are compensated to reach K neighbors, but it does not state the sampling distribution, whether the sample is fixed or redrawn per epoch or iteration, the random seed, or the compensation strategy. Because the accuracy results in Table 4 and the memory results in Table 7 both rely on this sampling, and because Table 6 varies K without redrawing neighbor sets, the reported K=4 result may reflect a single favorable draw rather than a robust property of the method.
  4. [§4.2] The narrative contradicts the reported results. The text says 'Our STAHGNet consistently outperforms most baseline methods in all datasets, except for PeMSD4 and PeMSD8. In PeMSD4, the lowest MAPE is owned by RGSL, and the RMSE of STAHGNet is slightly higher (8.4%) than DSTAGNN,' but Table 4 shows STAHGNet with the lowest MAPE on PeMSD4 (12.35 vs RGSL 12.80) and the lowest RMSE (29.62 vs DSTAGNN 30.86). This internal inconsistency makes the central performance claim ambiguous and must be resolved.
  5. [§4.1.3, Table 4] The paper states that each experiment was repeated five times and a paired t-test was used, but Table 4 reports only point estimates without standard deviations, confidence intervals, or p-values. Without these, the bolded 'statistically significant' claims are unverifiable, especially for small differences such as 0.1–0.5 in MAE. Please report the variance or statistical test results for the key comparisons.
  6. [§3.2.1, Eqs. (1)-(2)] The feature engineering can divide by zero. Equation (2) divides by x(i−1), which is zero for zero-flow time steps, and Eq. (1) divides by max(x) − min(x), which is zero for constant subseries. Traffic flow data commonly contain zero or constant intervals, so the preprocessing must specify how these cases are handled (e.g., epsilon smoothing or a fallback value). Without this, the model cannot be trained reliably on the described features.
minor comments (5)
  1. [§3.2.1 title] The title 'Automative Feature Engineering' contains a typo; it should be 'Automatic' or 'Automated Feature Engineering.'
  2. [§4.1.3] The phrase 'we used the evaluation protocol outlined in [39] to split all to split all datasets' has a duplicated clause and should be corrected.
  3. [§4.3, Table 5] Table 5 is very difficult to read because the metric values for w=11 and w=16 are interleaved with the parameter rows; please reformat the table so that each configuration is clearly separated.
  4. [§4.2] The sentence 'earlier graph models (i.e., STG2Seq, STGCN, STSGCN) used a GCN module ... which may explain their relatively worse performance compared to STSGCN' is confusing because STSGCN is included in the set of earlier graph models; please rephrase.
  5. [Reference [15]] The title of reference [15] uses 'multi-variates' instead of 'multivariate.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the model is trained and validated on disjoint splits, with no parameter fitted to test targets and no load-bearing self-citation chain.

full rationale

STAHGNet is an empirical deep-learning paper: the claimed results are obtained by training the recurrent/graph architecture on 60% training splits and measuring MAE/RMSE/MAPE on 20% test splits following the protocol of STSGCN [39], so the headline numbers are not constructed from the target values. Hyperparameters K, w, and H are selected by grid search on the validation set (Section 4.1.3, Table 5), which is standard model selection rather than fitting to the reported test result. The efficiency claim reduces to the complexity comparison from full-graph storage O(B x N x w) to sampled-neighbor storage O(B x K x w) stated in Section 3.2.1, an arithmetic bound rather than a fitted result. The equations defining the cell (Eq. 3), spatial aggregation (Eq. 4), temporal attention (Eq. 5), message passing (Eq. 6), CTG adjacency (Eqs. 7-8), and smooth-L1 loss (Eq. 9) are architectural definitions; none of them is defined in terms of the prediction target or of the reported benchmark numbers. The relevant citations (e.g., [18], [46], [47,48]) support standard components such as spatial-temporal attention and learned adjacency, and I find no author self-citation chain carrying a load-bearing uniqueness or forced-choice argument. The underspecification of the random neighbor sampling and compensation rule noted in Section 3.2.1 is a reproducibility concern, not a circularity one, since the sampled-neighbor subset is not itself the quantity being predicted. I therefore find no step that reduces by construction to its own inputs and assign score 0.

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

The central claim rests on a standard deep learning training pipeline plus several domain-specific assumptions: sampled neighbors are sufficient, missing values are handled silently, the CTG matrix is well-defined, and low-degree nodes are compensated without explanation. The free parameters are mostly standard hyperparameters, but the gamma schedule is an ad hoc choice that is also mathematically undefined at the boundary. No new physical or conceptual entities are introduced.

free parameters (5)
  • K, number of sampled neighbor nodes = 4 on PeMSD3/4/8; 6 on PeMSD7
    Tuned by grid search over [2,4,6,8] on the validation split; directly controls the efficiency-accuracy tradeoff (ablation in Table 6).
  • w, historical sliding window size = 11 or 16 depending on dataset
    Grid-searched over [11,16]; determines the input sequence length and therefore the receptive field.
  • H, neighbor sampling hop number = 1 or 2 depending on dataset
    Grid-searched over [1,2,3]; controls how many hops of neighbors are aggregated.
  • D, hidden dimension = 64
    Set by hand; capacity of all LSTM-style and attention representations.
  • gamma attenuation schedule = gamma(t)=1/(w-t)
    Hand-chosen decaying weights for CTG aggregation; undefined at t=w and not justified by experiments.
assumptions (5)
  • domain assumption The road network can be represented as a fixed graph G=(V,E), and a random subset of K neighbors carries the information needed for prediction.
    Introduced in Section 3.2.1; the efficiency claim rests on this, but no analysis quantifies information loss from sampling.
  • domain assumption Missing traffic values (up to 3.18%) can be imputed by an unspecified method without affecting the comparison.
    Section 4.1.1 reports missing ratios but never describes imputation; all models see the same preprocessed data, but the preprocessing is unstated.
  • standard math The attention and LSTM update equations (Eq. 3-5) are trainable by backpropagation and standard optimizers.
    Invoked implicitly in Section 3.2.2; these are standard deep learning primitives.
  • ad hoc to paper The adjacency matrix At in Eq. (8) is a valid similarity matrix, despite the equation being syntactically incomplete and lacking normalization.
    Section 3.2.3; as written, At = sigma(log(theta/(1-theta))) has no row normalization and the text mentions softmax before showing a sigmoid, so the operation is not well-defined.
  • ad hoc to paper The compensation strategy for nodes with fewer than K neighbors preserves the graph semantics.
    Section 3.2.1 says low-degree nodes are 'compensated' to satisfy K but does not say whether duplicates, zero-padding, or pseudo-nodes are used.

how reviews work

0 comments
Cite this review

Pith. "Pith review of STAHGNet: Modeling Hybrid-grained Heterogenous Dependency Efficiently for Traffic Prediction." pith.science (2026). https://pith.science/paper/ZHIF3JYG

@misc{pith2026241217524,
  author       = {Pith},
  title        = {Pith review of: STAHGNet: Modeling Hybrid-grained Heterogenous Dependency Efficiently for Traffic Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZHIF3JYG}},
  note         = {Machine review of arXiv:2412.17524}
}
read the original abstract

Traffic flow prediction plays a critical role in the intelligent transportation system, and it is also a challenging task because of the underlying complex Spatio-temporal patterns and heterogeneities evolving across time. However, most present works mostly concentrate on solely capturing Spatial-temporal dependency or extracting implicit similarity graphs, but the hybrid-granularity evolution is ignored in their modeling process. In this paper, we proposed a novel data-driven end-to-end framework, named Spatio-Temporal Aware Hybrid Graph Network (STAHGNet), to couple the hybrid-grained heterogeneous correlations in series simultaneously through an elaborately Hybrid Graph Attention Module (HGAT) and Coarse-granularity Temporal Graph (CTG) generator. Furthermore, an automotive feature engineering with domain knowledge and a random neighbor sampling strategy is utilized to improve efficiency and reduce computational complexity. The MAE, RMSE, and MAPE are used for evaluation metrics. Tested on four real-life datasets, our proposal outperforms eight classical baselines and four state-of-the-art (SOTA) methods (e.g., MAE 14.82 on PeMSD3; MAE 18.92 on PeMSD4). Besides, extensive experiments and visualizations verify the effectiveness of each component in STAHGNet. In terms of computational cost, STAHGNet saves at least four times the space compared to the previous SOTA models. The proposed model will be beneficial for more efficient TFP as well as intelligent transport system construction.

Figures

Figures reproduced from arXiv: 2412.17524 by the authors.

Figure 3
Figure 3. Specifically, at the t time-stamp, the X(t) and G(t) are inputted into the first cell in their own encoder, respectively. For example, (x (t) , x(t) ′, x(t)′′) is mapped into a high-dimension space to generate a representation vector hˆ(t) with size D. A state vector c (1) ∈ R 1×D is randomly initialized at the first time slot and keeps updating through the sequence. The spatial-temporal aware representation r (t−1)… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 38 canonical work pages

  1. [1]

    arXiv preprint arXiv:2106.09373 (2021)

    Yang, S.B., Guo, C., Hu, J., Tang, J., Yang, B.: Unsupervised path representa- tion learning with curriculum negative sampling. arXiv preprint arXiv:2106.09373 (2021)

    arXiv 2021

  2. [2]

    Transportation research

    Smith, B.L., Williams, B.M., Oswald, R.K.: Comparison of parametric and non- parametric models for traffic flow forecasting. Transportation research. Part C 10C(4), 303–321 (2002) 22

    work page 2002

  3. [3]

    Okutani, I., Stephanedes, Y.J.: Dynamic prediction of traffic volume through kalman filtering 18(1), 1–11 (1984)

    work page 1984

  4. [4]

    Transportation Research Part C Emerging Technologies 4(5), 307–318 (1996)

    Voort, M., Dougherty, M., Watson, S.: Combining kohonen maps with arima time series models to forecast traffic flow. Transportation Research Part C Emerging Technologies 4(5), 307–318 (1996)

    work page 1996

  5. [5]

    Transportation Research Part C 62, 21–34 (2016)

    Cai, P., Wang, Y., Lu, G., Chen, P., Ding, C., Sun, J.: A spatiotemporal cor- relative k-nearest neighbor model for short-term traffic multistep forecasting. Transportation Research Part C 62, 21–34 (2016)

    work page 2016

  6. [6]

    Applied mathematical modelling (3), 35 (2011)

    Hong, W.C., Dong, Y., Zheng, F., Lai, C.Y.: Forecasting urban traffic flow by svr with continuous aco. Applied mathematical modelling (3), 35 (2011)

    work page 2011

  7. [7]

    Pattern recognition 77, 354–377 (2018)

    Gu, J., Wang, Z., Kuen, J., Ma, L., Shahroudy, A., Shuai, B., Liu, T., Wang, X., Wang, G., Cai, J., et al.: Recent advances in convolutional neural networks. Pattern recognition 77, 354–377 (2018)

    work page 2018

  8. [8]

    IEEE transactions on Signal Processing 45(11), 2673–2681 (1997)

    Schuster, M., Paliwal, K.K.: Bidirectional recurrent neural networks. IEEE transactions on Signal Processing 45(11), 2673–2681 (1997)

    work page 1997

Show all 52 references
  1. [9]

    Advances in neural information processing systems 30 (2017)

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, L., Polosukhin, I.: Attention is all you need. Advances in neural information processing systems 30 (2017)

    2017

  2. [10]

    arXiv preprint arXiv:1709.04875 (2017)

    Yu, B., Yin, H., Zhu, Z.: Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. arXiv preprint arXiv:1709.04875 (2017)

    2017 arXiv

  3. [11]

    IEEE Transactions on Intelligent Transportation Systems 21(9), 3848–3858 (2019)

    Zhao, L., Song, Y., Zhang, C., Liu, Y., Wang, P., Lin, T., Deng, M., Li, H.: T-gcn: A temporal graph convolutional network for traffic prediction. IEEE Transactions on Intelligent Transportation Systems 21(9), 3848–3858 (2019)

    2019

  4. [12]

    IEEE transactions on neural networks20(1), 61–80 (2008)

    Scarselli, F., Gori, M., Tsoi, A.C., Hagenbuchner, M., Monfardini, G.: The graph neural network model. IEEE transactions on neural networks20(1), 61–80 (2008)

    2008

  5. [13]

    arXiv preprint arXiv:1905.10069 (2019)

    Bai, L., Yao, L., Kanhere, S., Wang, X., Sheng, Q., et al.: Stg2seq: Spatial- temporal graph to sequence model for multi-step passenger demand forecasting. arXiv preprint arXiv:1905.10069 (2019)

    2019 arXiv

  6. [14]

    Kazemi, S.M., Goel, R., Jain, K., Kobyzev, I., Sethi, A., Forsyth, P., Poupart, P.: Representation learning for dynamic graphs: A survey. J. Mach. Learn. Res. 21(70), 1–73 (2020)

    2020

  7. [15]

    arXiv preprint arXiv:2210.06126 (2022)

    Yu, H., Li, T., Yu, W., Li, J., Huang, Y., Wang, L., Liu, A.: Regularized graph structure learning with semantic knowledge for multi-variates time-series forecasting. arXiv preprint arXiv:2210.06126 (2022)

    2022 arXiv

  8. [16]

    arXiv preprint arXiv:1803.01271 (2018)

    Bai, S., Kolter, J.Z., Koltun, V.: An empirical evaluation of generic convolutional 23 and recurrent networks for sequence modeling. arXiv preprint arXiv:1803.01271 (2018)

    2018 arXiv

  9. [17]

    In: Proceedings of the 28th ACM Interna- tional Conference on Information and Knowledge Management, pp

    Huang, S., Wang, D., Wu, X., Tang, A.: Dsanet: Dual self-attention network for multivariate time series forecasting. In: Proceedings of the 28th ACM Interna- tional Conference on Information and Knowledge Management, pp. 2129–2132 (2019)

    2019

  10. [18]

    In: International Conference on Machine Learning, pp

    Lan, S., Ma, Y., Huang, W., Wang, W., Yang, H., Li, P.: Dstagnn: Dynamic spatial-temporal aware graph neural network for traffic flow forecasting. In: International Conference on Machine Learning, pp. 11906–11917 (2022). PMLR

    2022

  11. [19]

    arXiv (2022)

    Cirstea, R.-G., Yang, B., Guo, C., Kieu, T., Pan, S.: Towards Spatio-Temporal Aware Traffic Time Series Forecasting–Full Version. arXiv (2022). https://doi. org/10.48550/ARXIV.2203.15737 . https://arxiv.org/abs/2203.15737

    2022 doi arXiv

  12. [20]

    Modeling financial time series with S-PLUS ®, 385–429 (2006)

    Zivot, E., Wang, J.: Vector autoregressive models for multivariate time series. Modeling financial time series with S-PLUS ®, 385–429 (2006)

    2006

  13. [21]

    In: Intelligent Vehicles Symposium (2004)

    Vanajakshi, L., Rilett, L.R.: A comparison of the performance of artificial neural networks and support vector machines for the prediction of traffic speed. In: Intelligent Vehicles Symposium (2004)

    2004

  14. [22]

    In: 2018 IEEE 7th Data Driven Control and Learning Systems Conference (DDCLS), pp

    Dong, X., Lei, T., Jin, S., Hou, Z.: Short-term traffic flow prediction based on xgboost. In: 2018 IEEE 7th Data Driven Control and Learning Systems Conference (DDCLS), pp. 854–859 (2018). IEEE

    2018

  15. [23]

    Neural computation 9(8), 1735–1780 (1997)

    Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural computation 9(8), 1735–1780 (1997)

    1997

  16. [24]

    arXiv preprint arXiv:1412.3555 (2014)

    Chung, J., Gulcehre, C., Cho, K., Bengio, Y.: Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555 (2014)

    2014 arXiv

  17. [25]

    Advances in neural information processing systems 27 (2014)

    Sutskever, I., Vinyals, O., Le, Q.V.: Sequence to sequence learning with neural networks. Advances in neural information processing systems 27 (2014)

    2014

  18. [26]

    Advances in neural information processing systems 28 (2015)

    Shi, X., Chen, Z., Wang, H., Yeung, D.-Y., Wong, W.-K., Woo, W.-c.: Convolu- tional lstm network: A machine learning approach for precipitation nowcasting. Advances in neural information processing systems 28 (2015)

    2015

  19. [27]

    In: Proceedings of the 24th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, pp

    Zhang, J., Zheng, Y., Qi, D., Li, R., Yi, X.: Dnn-based prediction model for spatio- temporal data. In: Proceedings of the 24th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, pp. 1–4 (2016)

    2016

  20. [28]

    In: Thirty-first AAAI Conference on Artificial Intelligence 24 (2017)

    Zhang, J., Zheng, Y., Qi, D.: Deep spatio-temporal residual networks for citywide crowd flows prediction. In: Thirty-first AAAI Conference on Artificial Intelligence 24 (2017)

    2017

  21. [29]

    In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

    Yao, H., Tang, X., Wei, H., Zheng, G., Li, Z.: Revisiting spatial-temporal sim- ilarity: A deep learning framework for traffic prediction. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 33, pp. 5668–5675 (2019)

    2019

  22. [30]

    In: 2023 IEEE 39th International Conference on Data Engineering (ICDE), pp

    Fang, Y., Qin, Y., Luo, H., Zhao, F., Xu, B., Zeng, L., Wang, C.: When spatio- temporal meet wavelets: Disentangled traffic forecasting via efficient spectral graph attention networks. In: 2023 IEEE 39th International Conference on Data Engineering (ICDE), pp. 517–529 (2023). IEEE

    2023

  23. [31]

    In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

    Zhou, H., Zhang, S., Peng, J., Zhang, S., Li, J., Xiong, H., Zhang, W.: Informer: Beyond efficient transformer for long sequence time-series forecasting. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 11106–11115 (2021)

    2021

  24. [32]

    In: ICCV, vol

    Goldberger, J., Gordon, S., Greenspan, H., et al.: An efficient image similar- ity measure based on approximations of kl-divergence between two gaussian mixtures. In: ICCV, vol. 3, pp. 487–493 (2003)

    2003

  25. [33]

    arXiv preprint arXiv:2001.04451 (2020)

    Kitaev, N., Kaiser, L., Levskaya, A.: Reformer: The efficient transformer. arXiv preprint arXiv:2001.04451 (2020)

    2020 arXiv

  26. [34]

    In: Proceed- ings of the 39th International Conference on Machine Learning

    Zhou, T., Ma, Z., Wen, Q., Wang, X., Sun, L., Jin, R.: FEDformer: Frequency enhanced decomposed transformer for long-term series forecasting. In: Proceed- ings of the 39th International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 162, pp. 272...

    2022

  27. [35]

    In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

    Jin, D., Shi, J., Wang, R., Li, Y., Huang, Y., Yang, Y.-B.: Trafformer: Unify time and space in traffic prediction. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, pp. 8114–8122 (2023)

    2023

  28. [36]

    arXiv preprint arXiv:1312.6203 (2013)

    Bruna, J., Zaremba, W., Szlam, A., LeCun, Y.: Spectral networks and locally connected networks on graphs. arXiv preprint arXiv:1312.6203 (2013)

    2013 arXiv

  29. [37]

    Advances in neural information processing systems 29 (2016)

    Defferrard, M., Bresson, X., Vandergheynst, P.: Convolutional neural networks on graphs with fast localized spectral filtering. Advances in neural information processing systems 29 (2016)

    2016

  30. [38]

    IEEE Transactions on Neural Networks 20(3), 498–511 (2009)

    Micheli, A.: Neural network for graphs: A contextual constructive approach. IEEE Transactions on Neural Networks 20(3), 498–511 (2009)

    2009

  31. [39]

    In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

    Song, C., Lin, Y., Guo, S., Wan, H.: Spatial-temporal synchronous graph convolu- tional networks: A new framework for spatial-temporal network data forecasting. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, pp. 914–921 (2020) 25

    2020

  32. [40]

    stat 1050, 20 (2017)

    Velickovic, P., Cucurull, G., Casanova, A., Romero, A., Lio, P., Bengio, Y.: Graph attention networks. stat 1050, 20 (2017)

    2017

  33. [41]

    Neurocomputing 410, 387–393 (2020)

    Guo, G., Yuan, W.: Short-term traffic speed forecasting based on graph attention temporal convolutional networks. Neurocomputing 410, 387–393 (2020)

    2020

  34. [42]

    IEEE Access 7, 166246–166256 (2019)

    Zhang, C., James, J., Liu, Y.: Spatial-temporal graph attention networks: A deep learning approach for traffic forecasting. IEEE Access 7, 166246–166256 (2019)

    2019

  35. [43]

    IEEE Transactions on Intelligent Transportation Systems (2022)

    Huang, J., Luo, K., Cao, L., Wen, Y., Zhong, S.: Learning multiaspect traffic couplings by multirelational graph attention networks for traffic prediction. IEEE Transactions on Intelligent Transportation Systems (2022)

    2022

  36. [44]

    In: Proceedings of the AAAI Conference on Artificial Intelligence, vol

    Li, M., Zhu, Z.: Spatial-temporal fusion graph neural networks for traffic flow forecasting. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 4189–4196 (2021)

    2021

  37. [45]

    In: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp

    Ye, J., Liu, Z., Du, B., Sun, L., Li, W., Fu, Y., Xiong, H.: Learning the evolu- tionary and multi-scale graph structure for multivariate time series forecasting. In: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 2296–2306 (2022)

    2022

  38. [46]

    Elsayed, S., Thyssens, D., Rashed, A., Jomaa, H.S., Schmidt-Thieme, L.: Do we really need deep learning models for time series forecasting? arXiv preprint arXiv:2101.02118 (2021)

    2021 arXiv

  39. [47]

    arXiv preprint arXiv:1906.00121 (2019)

    Wu, Z., Pan, S., Long, G., Jiang, J., Zhang, C.: Graph wavenet for deep spatial- temporal graph modeling. arXiv preprint arXiv:1906.00121 (2019)

    2019 arXiv

  40. [48]

    Advances in neural information processing systems 33, 17804–17815 (2020)

    Bai, L., Yao, L., Li, C., Wang, X., Wang, C.: Adaptive graph convolutional recur- rent network for traffic forecasting. Advances in neural information processing systems 33, 17804–17815 (2020)

    2020

  41. [49]

    In: ICNN, pp

    Irie, B., Miyake, S.: Capabilities of three-layered perceptrons. In: ICNN, pp. 641– 648 (1988)

    1988

  42. [50]

    The journal of machine learning research 15(1), 1929–1958 (2014)

    Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., Salakhutdinov, R.: Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research 15(1), 1929–1958 (2014)

    2014

  43. [51]

    Engineering Applications of Artificial Intelligence 124, 106615 (2023)

    Zhao, W., Zhang, S., Zhou, B., Wang, B.: Multi-spatio-temporal fusion graph recurrent network for traffic forecasting. Engineering Applications of Artificial Intelligence 124, 106615 (2023)

    2023

  44. [52]

    Advances in neural information processing 26 systems 32 (2019) 27

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al.: Pytorch: An imperative style, high- performance deep learning library. Advances in neural information processing 26 systems 32 (2019) 27

    2019

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.