Pith. sign in

REVIEW 4 major objections 6 minor 74 references

Unveiling the Inflexibility of Adaptive Embedding in Traffic Forecasting

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Trainable node embeddings are the main source of inflexibility in traffic forecasting; replacing them with PCA-derived embeddings restores cross-year accuracy and enables zero-shot city transfer.

desk verdict Cross-year setup is confounded because PCA is fit on target-year data, but the zero-shot city-transfer experiments and the diagnosis of stale adaptive embeddings are real and deserve review. read the letter →

arxiv 2411.11448 v1 pith:PD7LQ5ER submitted 2024-11-18 cs.LG cs.AI

Hongjun Wang , Jiyuan Chen , Lingyu Zhang , Renhe Jiang , Xuan Song This is my paper · ORCID
classification cs.LGcs.AI
keywords trafficforecastingadaptiveembeddingPCAspatiotemporalgraphneuralnetworksdistributionshiftzero-shottransfercross-citygeneralization
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

Current spatiotemporal traffic forecasting models—both graph neural networks and Transformers—degrade sharply when tested a year after training or on a different city. This paper argues that the degradation is caused by the trainable adaptive embedding layer that encodes node identities, not by the temporal architecture: an LSTM with no such embeddings stays stable across years. The proposed remedy replaces trainable embeddings with PCA embeddings computed from the data via the projection matrix $P$, which is learned once and reused at test time to map new inputs into the same low-dimensional space. With this substitution, cross-year error drops substantially (e.g., STID on PEMS03 average MAE from 33.99 to 9.18) and zero-shot cross-city prediction becomes usable without retraining. If correct, this identifies a single swap-able component as the bottleneck and offers a training-free path to adapting traffic models to evolving urban environments.

What carries the argument

The PCA embedding is the central object: for each day $d$, the traffic matrix $Z_d \in \mathbb{R}^{N \times T}$ (nodes by daily time slots) is projected onto the principal subspace learned from training data, $E_d = Z_d P$, and the final node embedding is the average over training days, $E_{\mathrm{pca}} = \frac{1}{D}\sum_d Z_d P$. At test time the same projection matrix $P$ maps new data into the same space, allowing the model to build a fresh graph and process node sets of different sizes without retraining. This machinery directly counters the three diagnosed limitations of adaptive embeddings: it gives inductive capacity because the embedding is recomputed from data, balanced distinguishability because PCA truncation retains only major variance directions, and transferability because the same basis serves different cities. The choice of the number of principal components acts as a knob controlling spatial distinctiveness, with experiments showing optimal values around 4–8 components for STID and STAEformer on PEMS03.

What would settle it

Measure the fraction of total variance in the target-year or target-city traffic data that is preserved when projected onto the source PCA subspace. If that fraction drops sharply (for instance, below the level at which the top components explain the training data), then the cross-year and zero-shot gains should collapse, and the model should fall back to the degraded performance seen with trained embeddings. A concrete experiment: train on PEMS03, project PEMS04 or PEMS08 data onto the PEMS03 subspace, and correlate the explained-variance ratio with the observed MAE across different target years and cities.

Watch

Extended reading notes

Core claim

Adaptive embeddings—randomly initialized learnable node representations that are updated during training to distinguish locations with similar historical patterns—are the principal cause of inflexibility in spatiotemporal forecasting under spatial shift. When the urban environment shifts (a later year or a different city), these fixed embeddings carry outdated spatial biases, and models that rely on them collapse, whereas an LSTM without such embeddings maintains consistent error. Replacing adaptive embeddings with PCA embeddings, defined as $E_{\mathrm{pca}} = ZP$ where $Z$ is the day-sliced traffic feature matrix and $P$ is the PCA projection learned from the training data (including a 5% validation subset), largely fixes the problem: in-distribution accuracy is preserved or improved, cross-year average MAE drops sharply on four PEMS benchmarks, and zero-shot transfer between cities becomes feasible across several architectures. This is interpreted as evidence that the principal subspace of traffic patterns is sufficiently stable across years and cities to serve as a universal coordinate system for node identity.

Load-bearing premise

The entire cross-year and zero-shot benefit rests on the assumption that the PCA projection matrix learned from the training year remains a valid coordinate system for later-year or other-city traffic data; if the target data has variance in directions outside the source subspace, the projection discards exactly the information that distinguishes the new sensors.

Editorial extensions

If this is right

  • If the claim holds, traffic forecasting models can be trained once and then applied to a later year's data with substantially lower error, eliminating the need for yearly retraining.
  • Zero-shot cross-city deployment becomes practical: a model trained on one city's sensors can forecast traffic in another city without fine-tuning, as demonstrated on PEMS and LargeST pairs.
  • The adaptive embedding layer, rather than the graph or temporal architecture, is the common bottleneck; therefore, swapping in PCA embeddings should improve any ST-GNN or Transformer that uses trainable node embeddings.
  • PCA embeddings make the graph structure flexible between training and testing, so sensor additions, removals, or failures no longer require model modification—only a recomputation of the embedding from the new feature matrix.
  • In-distribution accuracy is not sacrificed for generalization; in several cases the frozen PCA embedding outperforms the trained adaptive embedding because it curbs overfitting and over-distinguishability.

Reading between the lines

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

  • A testable extension: if the source PCA subspace is a stable 'traffic manifold,' then the same projection matrix $P$ could be applied across many cities and years, opening the door to a shared embedding space for large-scale traffic foundation models—something the paper hints at but does not fully develop.
  • The paper's diagnosis suggests that fine-tuning adaptive embeddings is a stopgap: fine-tuning matches PCA on the tested benchmarks, but because it re-fits embeddings to the target distribution, it should be more sample-hungry than the training-free PCA projection, especially when the target has few labeled samples—an experiment the paper does not run.
  • A boundary condition for the method: the stability of the principal subspace across domains is assumed rather than proven; one could stress-test it with a city that has a fundamentally new transport mode or a radically different sensor layout, where the source subspace might fail to span the target variance.
  • The finding may generalize beyond traffic to other spatiotemporal prediction tasks (energy demand, crowd flow, disease spread) wherever node identities shift over time, but the subspace-stability assumption would need to be verified for each domain.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper argues that trainable adaptive node embeddings are a central cause of poor cross-year and cross-city generalization in traffic forecasting models. To support this, it introduces an alternative embedding computed by PCA: E_pca = Z · P, where P is a projection matrix obtained from the input statistics, and reports that replacing learnable embeddings with PCA embeddings preserves in-distribution accuracy while yielding large cross-year gains (e.g., STID on PEMS03 2018→2019, average MAE 33.99→9.18) and usable zero-shot cross-city transfer (Table I). The paper also introduces four new cross-year PEMS benchmarks and compares PCA embeddings against zero-embedding and fine-tuning strategies.

Significance. If the central claims held, the paper would provide a simple, parameter-light remedy for a recognized practical problem: adaptive embeddings that cannot adapt when sensor networks or urban spatial relationships change. The cross-year benchmark itself is a useful contribution, and the idea of deriving node embeddings from input statistics rather than from learned parameters is worth further study. The paper is also careful to include zero-embedding and fine-tuning comparisons in Table III, which helps isolate the role of the embedding. However, the main experimental evidence is currently confounded by the use of target-year data in fitting the PCA projection, and several reporting choices (single-seed results, per-dataset model selection, unexplained MAPE values) prevent the claims from being accepted as stated. The potential significance is real, but the evidence as presented is not yet load-bearing.

major comments (4)
  1. [Section IV, Section III-C, Eqs. (1)-(2)] The cross-year experiments in Tables IV and V are not a fair test of the claim that PCA embeddings enable training-free adaptation. Section IV states: "To prevent information leakage, we utilized 5% of the data from the second year for PCA dimensionality reduction to generate embeddings, and then tested on the remaining 95%." That means the projection matrix P is fit on a slice of the target-year distribution, whereas the frozen baselines see no target-year data. Eq. (2) and the surrounding text say that "the same PCA projection matrix, P" from training is used at test time, which is inconsistent with the experimental setup. The reported gains (e.g., STID on PEMS03: MAE 33.99→9.18) could therefore be driven by transductive access to the target-year covariance structure rather than by a general advantage of PCA embeddings. The authors must rerun Tables IV and V with P computed only from source-year training data (and possibly source-year validation data), and state explicitly whether any target-year data are used in fitting P. If the gains disappear, the central claim about adaptive embeddings being the main source of inflexibility is not supported by the paper's primary evidence.
  2. [Table III] Table III selects a different model for each dataset (AGCRN for PEMS03, GWNet for PEMS04, STID for PEMS07, STAEformer for PEMS08) without stating the selection protocol. If the PCA-embedding comparison is meant to show a general effect, the paper should report results for all evaluated models on all four datasets, or at least specify an a priori rule for choosing the displayed model. As it stands, the reader cannot tell whether the reported PCA-vs-baseline and PCA-vs-fine-tune comparisons are representative or cherry-picked. This matters because the fine-tuning comparison in Table III is the cleanest controlled evidence in the paper, and it should not be limited to favorable examples.
  3. [Tables IV and V, MAPE definition] The MAPE values in Tables IV and V are not credible as reported. For example, PEMS03 2018→2019 baseline MAPEs range from 753% to 1329% while MAEs are in the 20-50 range (e.g., AGCRN average MAE 28.04 and MAPE 820.77%). Such values are impossible under the standard MAPE definition unless the data contain many near-zero ground-truth values and the error is not masked consistently. The paper says zero values are excluded "as they represent noisy data," but the exact masking rule, the proportion of masked samples, and the MAPE formula are not given. The "IMPROVE" rows (e.g., 90.5% MAPE improvement on PEMS03) are largely driven by these inflated baseline MAPEs. The authors should report MAPE on a common, clearly defined subset of valid samples and also report error distributions or additional scale-free metrics that are robust to near-zero values.
  4. [All experimental tables] No results are averaged over multiple random seeds, and no standard deviations or confidence intervals are reported. The in-distribution comparisons in Figure 5 and the cross-year comparisons in Tables IV-V show differences that are often small relative to the typical run-to-run variance of these models (e.g., many MAE differences below 1.0 in Figure 5). For the central quantitative claims, the paper needs at least three seeds per configuration and a statement of how the reported numbers are selected. This is especially important because Section IV says the authors "performed multiple rounds of parameter tuning to achieve optimal results," which raises the risk of selecting favorable runs.
minor comments (6)
  1. [Abstract and Section IV] The claim that the method is "training-free" is overstated: PCA itself is a fitted projection, and in the current experiments it is fit on 5% of the target year. The abstract and Section I should distinguish between source-only PCA (genuinely zero-shot) and target-adapted PCA (transductive).
  2. [Section III-C, Eq. (2)] The notation is inconsistent: Eq. (2) defines E_test using Z_val, but the text says P is the projection from training data. If the intended protocol is to reuse P from training, the test-time equation should show this, and the validation subset should not be introduced as the source of P for the cross-year experiments.
  3. [Section I and Section IV] The introduction contains duplicated paragraphs: the paragraph beginning "In this paper, we investigate the core limitations" appears twice with only minor rewording. Please remove the redundancy.
  4. [Section IV, dataset description] The paper states that the code is "released in code," but no URL is given. Since the new cross-year benchmarks are a central contribution, the datasets and code should be made available with a stable link, and a data card should describe how the PEMS sensor sets were matched across years.
  5. [Table I] For the cross-city zero-shot experiments, please clarify whether the PCA projection matrix is computed from the source city only or from a subset of the target city. The text says "projection matrices W derived from PEMS03 and PEMS07" are applied to PEMS04/PEMS08, but the LargeST description is less explicit; a direct statement for each block of Table I would remove ambiguity.
  6. [Section IV-F, Figure 7] The optimal number of principal components is selected on the test set (PEMS03) by grid search. Please state how the number of components and the variance threshold θ are chosen in all other experiments, and whether the selection is made on validation data only, to avoid test-set tuning.

Circularity Check

1 steps flagged · score 4.0 of 10

Cross-year 'training-free' gains are partially confounded: PCA projection P is fitted to 5% of the target-year data before testing on the remaining 95%.

  1. fitted input called prediction [The cross-year setup is described in Section IV (RQ3), and Eq. (2) is defined in Section III-C.]
    "To prevent information leakage, we utilized 5% of the data from the second year for PCA dimensionality reduction to generate embeddings, and then tested on the remaining 95%."

    Cross-year evaluation is presented as evidence that PCA embeddings generalize without retraining, but the P in Eq. (2) is not the training-year projection in this experiment: it is estimated from 5% of the second-year data. Thus the model sees target-year input covariance before forecasting the remaining 95%, while AGCRN, STID, GWNet, and other baselines use frozen adaptive embeddings with no second-year information. The large MAE drops are therefore partly a transductive effect of fitting the coordinate system to the target distribution, not a pure comparison of PCA embeddings versus adaptive embeddings for out-of-distribution generalization.

full rationale

The central proposal is not circular in the strong sense: PCA embeddings are computed from input statistics, not from prediction targets, and the cross-city zero-shot experiments (Table I) derive the projection from the source city and apply it to the target city, which is independent support. However, the cross-year evidence (Tables IV and V) contains a concrete fitted-input confound. Section IV states that PCA dimensionality reduction for the second year uses 5% of that year's data before testing on the remaining 95%. Equation (2) then defines the test embedding as Z_val * P, so the projection P is a parameter fitted to the target-year input distribution. The frozen adaptive-embedding baselines receive no target-year data, so the comparison is not apples-to-apples: the reported gains (e.g., STID on PEMS03, MAE 33.99 to 9.18) demonstrate transductive adaptation, not the training-free inductive generalization claimed in the abstract and Section III. No load-bearing self-citation chain was found, and the cross-city experiments remain independent, so the score is 4 rather than higher.

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

The central claim rests on four premises the reader must accept without independent proof: daily periodicity of traffic, transferability of the PCA subspace across time and cities, LSTM as a valid spatial-shift control, and linear PCA retaining enough sensor-discriminative information. The only fitted quantities are the number of principal components, the variance threshold mentioned in the text, and the 5% target-data fraction used in cross-year experiments.

free parameters (3)
  • Number of principal components C = 4 for STID and 8 for STAEformer on PEMS03 (grid-searched)
    The paper searches over component counts in Fig 7 and tunes the count per model and dataset; this is a validation-fitted hyperparameter, not derived from the data alone.
  • Variance threshold theta = not specified in experiments
    Section III-C1 suggests keeping components until the explained variance fraction reaches theta, but the experiments instead grid-search counts, so theta remains an unspecified tuning target.
  • Validation subset fraction for PCA fitting = 5%
    Section IV states that 5% of the target-year data is used to compute PCA embeddings in cross-year experiments; this fraction is chosen by hand and affects how much target information leaks into the embedding.
assumptions (4)
  • domain assumption Traffic data has stable daily periodicity, allowing each day to be cut into T equal time slots and reshaped into Z in D x N x T.
    Invoked in Section III-C when defining the PCA input matrix; if periodicity is weak or the slot structure is wrong, the PCA embeddings lose meaning.
  • domain assumption The principal subspace of traffic features is stable across years and cities, so a projection P trained on source data can be reused on target data.
    This is the load-bearing premise behind Eq. (2) and the zero-shot experiments in Table I.
  • domain assumption LSTM serves as a valid no-spatial control, so the degradation gap between LSTM and ST-GNNs isolates spatial shift rather than temporal or capacity differences.
    Figure 2 and Section I use LSTM's stable performance to attribute the degradation to spatial relationships.
  • domain assumption Linear PCA of raw input series preserves the sensor-discriminative information that adaptive embeddings provide.
    Section III-C1 claims PCA maintains balanced spatial representation while avoiding overfitting; this is asserted, not proven.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unveiling the Inflexibility of Adaptive Embedding in Traffic Forecasting." pith.science (2026). https://pith.science/paper/PD7LQ5ER

@misc{pith2026241111448,
  author       = {Pith},
  title        = {Pith review of: Unveiling the Inflexibility of Adaptive Embedding in Traffic Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PD7LQ5ER}},
  note         = {Machine review of arXiv:2411.11448}
}
read the original abstract

Spatiotemporal Graph Neural Networks (ST-GNNs) and Transformers have shown significant promise in traffic forecasting by effectively modeling temporal and spatial correlations. However, rapid urbanization in recent years has led to dynamic shifts in traffic patterns and travel demand, posing major challenges for accurate long-term traffic prediction. The generalization capability of ST-GNNs in extended temporal scenarios and cross-city applications remains largely unexplored. In this study, we evaluate state-of-the-art models on an extended traffic benchmark and observe substantial performance degradation in existing ST-GNNs over time, which we attribute to their limited inductive capabilities. Our analysis reveals that this degradation stems from an inability to adapt to evolving spatial relationships within urban environments. To address this limitation, we reconsider the design of adaptive embeddings and propose a Principal Component Analysis (PCA) embedding approach that enables models to adapt to new scenarios without retraining. We incorporate PCA embeddings into existing ST-GNN and Transformer architectures, achieving marked improvements in performance. Notably, PCA embeddings allow for flexibility in graph structures between training and testing, enabling models trained on one city to perform zero-shot predictions on other cities. This adaptability demonstrates the potential of PCA embeddings in enhancing the robustness and generalization of spatiotemporal models.

Figures

Figures reproduced from arXiv: 2411.11448 by the authors.

Figure 1
Figure 1. Spatiotemporal analysis of urban development patterns in Sacramento, California (2016-2018). The sequence demonstrates the progressive intensification of building density (shown in blue) and its implications for transportation demand modeling. Red circles highlight key areas of urban transformation, indicating the dynamic nature of land use changes and their potential impact on travel demand patterns [PITH_FULL_IMA… view at source ↗
Figure 2
Figure 2. We conducted a comparative analysis of LSTM against state-of-the-art (SOTA) models, including , STID [11], GWNet [12], AGCRN [13], MTGNN [14], TrendGCN [15], and STAEformer [16], assessing their performance on both in-distribution and out-of-distribution test datasets. In the in-distribution scenario, we utilized the original dataset, which had only a few weeks’ interval from the training set, while the out-of-distr… view at source ↗
Figure 3
Figure 3. We found that adaptive embedding restricts the model to perform inference on the same graph, which is unrealistic in dynamic traffic scenarios due to the continuous development of cities. We propose a novel testing-time adaptive strategy that requires no additional training, where ’fire’ indicates the optimization of target parameters and ’snowflake’ represents the freezing of model parameters. Specifically, we appl… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Performance comparison of STID and AGCRN models with and without adaptive embedding across four datasets (METR-LA, PEMS￾BAY, ER, and ETTm1). We observe that employing trainable adaptive embeddings results in excessive spatial distinguishability, leading to a per￾forman…
Figure 5
Figure 5. Figure 5: We conducted a comparative analysis of model performance on the PEMS benchmark using PCA embeddings versus learnable embeddings. The MAE results reveal that employing PCA (frozen) embeddings does not deteriorate training outcomes. Surprisingly, in certain instances, it…
Figure 6
Figure 6. Figure 6: Visualization of pairwise PCA-embedding alignment for traffic data: Each scatter plot illustrates traffic pattern comparisons across different PEMS datasets, where blue points represent source dataset embeddings and red points represent target dataset embeddings. The 4…
Figure 7
Figure 7. Figure 7: Component search results for the optimal number of principal components in STID and STAEformer when applied to the PEMS03 dataset are presented. The red dashed lines signify the baseline performance. The fluctuating MAE curves indicate that traditional methods are susc…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 43 canonical work pages

  1. [1]

    Discrete graph structure learning for fore- casting multiple time series,

    C. Shang, J. Chen, and J. Bi, “Discrete graph structure learning for fore- casting multiple time series,” in International Conference on Learning Representations, 2021

    work page 2021

  2. [2]

    Dstagnn: Dynamic spatial-temporal aware graph neural network for traffic flow forecasting,

    S. Lan, Y . Ma, W. Huang, W. Wang, H. Yang, and P. Li, “Dstagnn: Dynamic spatial-temporal aware graph neural network for traffic flow forecasting,” in International conference on machine learning . PMLR, 2022, pp. 11 906–11 917

    2022

  3. [3]

    Learning to remember patterns: Pattern matching memory networks for traffic forecasting,

    H. Lee, S. Jin, H. Chu, H. Lim, and S. Ko, “Learning to remember patterns: Pattern matching memory networks for traffic forecasting,” in International Conference on Learning Representations , 2022

    work page 2022

  4. [4]

    Spatio-temporal meta-graph learning for traffic forecasting,

    R. Jiang, Z. Wang, J. Yong, P. Jeph, Q. Chen, Y . Kobayashi, X. Song, S. Fukushima, and T. Suzumura, “Spatio-temporal meta-graph learning for traffic forecasting,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 7, 2023, pp. 8078–8086

    work page 2023

  5. [5]

    Deciphering spatio-temporal graph forecasting: A causal lens and treatment,

    Y . Xia, Y . Liang, H. Wen, X. Liu, K. Wang, Z. Zhou, and R. Zimmer- mann, “Deciphering spatio-temporal graph forecasting: A causal lens and treatment,” in Thirty-seventh Conference on Neural Information Processing Systems, 2023

    work page 2023

  6. [6]

    Largest: A benchmark dataset for large- scale traffic forecasting,

    X. Liu, Y . Xia, Y . Liang, J. Hu, Y . Wang, L. Bai, C. Huang, Z. Liu, B. Hooi, and R. Zimmermann, “Largest: A benchmark dataset for large- scale traffic forecasting,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  7. [7]

    Learning social meta-knowledge for nowcasting human mobility in disaster,

    R. Jiang, Z. Wang, Y . Tao, C. Yang, X. Song, R. Shibasaki, S.-C. Chen, and M.-L. Shyu, “Learning social meta-knowledge for nowcasting human mobility in disaster,” in Proceedings of the ACM Web Conference 2023, 2023, pp. 2655–2665

    work page 2023

  8. [8]

    Curb-gan: Conditional urban traffic estimation through spatio-temporal generative adversarial networks,

    Y . Zhang, Y . Li, X. Zhou, X. Kong, and J. Luo, “Curb-gan: Conditional urban traffic estimation through spatio-temporal generative adversarial networks,” in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , 2020, pp. 842– 852

    work page 2020

Show all 74 references
  1. [9]

    Maintaining the status quo: Capturing invariant relations for ood spatiotemporal learning,

    Z. Zhou, Q. Huang, K. Yang, K. Wang, X. Wang, Y . Zhang, Y . Liang, and Y . Wang, “Maintaining the status quo: Capturing invariant relations for ood spatiotemporal learning,” in Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023, pp. 3603– 3614

    2023

  2. [10]

    Self-supervised deconfounding against spatio-temporal shifts: Theory and modeling,

    J. Ji, W. Zhang, J. Wang, Y . He, and C. Huang, “Self-supervised deconfounding against spatio-temporal shifts: Theory and modeling,” arXiv preprint arXiv:2311.12472 , 2023

    2023 arXiv

  3. [11]

    Spatial-temporal identity: A simple yet effective baseline for multivariate time series forecasting,

    Z. Shao, Z. Zhang, F. Wang, W. Wei, and Y . Xu, “Spatial-temporal identity: A simple yet effective baseline for multivariate time series forecasting,” in Proceedings of the 31st ACM International Conference on Information & Knowledge Management , 2022, pp. 4454–4458

    2022

  4. [12]

    Graph wavenet for deep spatial-temporal graph modeling,

    Z. Wu, S. Pan, G. Long, J. Jiang, and C. Zhang, “Graph wavenet for deep spatial-temporal graph modeling,” in Proceedings of the 28th International Joint Conference on Artificial Intelligence, 2019, pp. 1907– 1913

    2019

  5. [13]

    Adaptive graph convolutional recurrent network for traffic forecasting,

    L. Bai, L. Yao, C. Li, X. Wang, and C. Wang, “Adaptive graph convolutional recurrent network for traffic forecasting,” Advances in neural information processing systems, vol. 33, pp. 17 804–17 815, 2020

    2020

  6. [15]

    Enhancing the ro- bustness via adversarial learning and joint spatial-temporal embeddings in traffic forecasting,

    J. Jiang, B. Wu, L. Chen, K. Zhang, and S. Kim, “Enhancing the ro- bustness via adversarial learning and joint spatial-temporal embeddings in traffic forecasting,” in Proceedings of the 32nd ACM International Conference on Information and Knowledge Management, 2023, pp. 987– 996

    2023

  7. [16]

    Spatio-temporal adaptive embedding makes vanilla transformer sota for traffic forecasting,

    H. Liu, Z. Dong, R. Jiang, J. Deng, J. Deng, Q. Chen, and X. Song, “Spatio-temporal adaptive embedding makes vanilla transformer sota for traffic forecasting,” in Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , 2023, pp. 4125–4129

    2023

  8. [17]

    Exploring progress in multivariate time series forecasting: Comprehensive benchmarking and heterogeneity analysis,

    Z. Shao, F. Wang, Y . Xu, W. Wei, C. Yu, Z. Zhang, D. Yao, G. Jin, X. Cao, G. Cong et al., “Exploring progress in multivariate time series forecasting: Comprehensive benchmarking and heterogeneity analysis,” arXiv preprint arXiv:2310.06119 , 2023

    arXiv

  9. [18]

    Attention based spatial- temporal graph convolutional networks for traffic flow forecasting,

    S. Guo, Y . Lin, N. Feng, C. Song, and H. Wan, “Attention based spatial- temporal graph convolutional networks for traffic flow forecasting,” in Proceedings of the AAAI conference on artificial intelligence , vol. 33, no. 01, 2019, pp. 922–929

    2019

  10. [19]

    Diffusion convolutional recurrent neural network: Data-driven traffic forecasting,

    Y . Li, R. Yu, C. Shahabi, and Y . Liu, “Diffusion convolutional recurrent neural network: Data-driven traffic forecasting,” in International Con- ference on Learning Representations , 2018

    2018

  11. [20]

    Dl-traff: Survey and benchmark of deep learning models for urban traffic prediction,

    R. Jiang, D. Yin, Z. Wang, Y . Wang, J. Deng, H. Liu, Z. Cai, J. Deng, X. Song, and R. Shibasaki, “Dl-traff: Survey and benchmark of deep learning models for urban traffic prediction,” in Proceedings of the 30th ACM international conference on information & knowledge managemen...

    2021

  12. [21]

    Freeway performance measurement system: mining loop detector data,

    C. Chen, K. Petty, A. Skabardonis, P. Varaiya, and Z. Jia, “Freeway performance measurement system: mining loop detector data,” Trans- portation research record, pp. 96–102, 2001

    2001

  13. [22]

    Spatial-temporal synchronous graph convolutional networks: A new framework for spatial-temporal network data forecasting,

    C. Song, Y . Lin, S. Guo, and H. Wan, “Spatial-temporal synchronous graph convolutional networks: A new framework for spatial-temporal network data forecasting,” in Proceedings of the AAAI Conference on Artificial Intelligence, 2020, pp. 914–921

    2020

  14. [23]

    Attention based spatial- temporal graph convolutional networks for traffic flow forecasting,

    S. Guo, Y . Lin, N. Feng, C. Song, and H. Wan, “Attention based spatial- temporal graph convolutional networks for traffic flow forecasting,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 33, no. 01, 2019, pp. 922–929

    2019

  15. [24]

    Vision transformers need registers,

    T. Darcet, M. Oquab, J. Mairal, and P. Bojanowski, “Vision transformers need registers,” arXiv preprint arXiv:2309.16588 , 2023

    2023 arXiv

  16. [25]

    A comprehensive survey on transfer learning,

    F. Zhuang, Z. Qi, K. Duan, D. Xi, Y . Zhu, H. Zhu, H. Xiong, and Q. He, “A comprehensive survey on transfer learning,” Proceedings of the IEEE, vol. 109, no. 1, pp. 43–76, 2020

    2020

  17. [26]

    Online learning: A compre- hensive survey,

    S. C. Hoi, D. Sahoo, J. Lu, and P. Zhao, “Online learning: A compre- hensive survey,” Neurocomputing, vol. 459, pp. 249–289, 2021

    2021

  18. [27]

    Principal component analysis,

    S. Wold, K. Esbensen, and P. Geladi, “Principal component analysis,” Chemometrics and intelligent laboratory systems , vol. 2, no. 1-3, pp. 37–52, 1987

    1987

  19. [28]

    A review of generalized zero-shot learning meth- ods,

    F. Pourpanah, M. Abdar, Y . Luo, X. Zhou, R. Wang, C. P. Lim, X.-Z. Wang, and Q. J. Wu, “A review of generalized zero-shot learning meth- ods,” IEEE transactions on pattern analysis and machine intelligence , vol. 45, no. 4, pp. 4051–4070, 2022

    2022

  20. [29]

    A computer movie simulating urban growth in the detroit region,

    W. R. Tobler, “A computer movie simulating urban growth in the detroit region,” Economic geography, vol. 46, no. sup1, pp. 234–240, 1970

    1970

  21. [30]

    Stg4traffic: A survey and bench- mark of spatial-temporal graph neural networks for traffic prediction,

    X. Luo, C. Zhu, D. Zhang, and Q. Li, “Stg4traffic: A survey and bench- mark of spatial-temporal graph neural networks for traffic prediction,” arXiv preprint arXiv:2307.00495 , 2023

    2023 arXiv

  22. [31]

    Spatio-temporal graph convolutional net- works: A deep learning framework for traffic forecasting,

    B. Yu, H. Yin, and Z. Zhu, “Spatio-temporal graph convolutional net- works: A deep learning framework for traffic forecasting,” in Proceed- ings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI-18, 2018, pp. 3634–3640

    2018

  23. [32]

    Decoupled dynamic spatial-temporal graph neural network for traffic forecasting,

    Z. Shao, Z. Zhang, W. Wei, F. Wang, Y . Xu, X. Cao, and C. S. Jensen, “Decoupled dynamic spatial-temporal graph neural network for traffic forecasting,” in Proceedings of the VLDB Endowment , 2022, pp. 2733– 2746

    2022

  24. [33]

    Coupled layer-wise graph convolution for transportation demand prediction,

    J. Ye, L. Sun, B. Du, Y . Fu, and H. Xiong, “Coupled layer-wise graph convolution for transportation demand prediction,” in AAAI, vol. 35, no. 5, 2021, pp. 4617–4625

    2021

  25. [34]

    Dynamic and multi- faceted spatio-temporal deep learning for traffic speed forecasting,

    L. Han, B. Du, L. Sun, Y . Fu, Y . Lv, and H. Xiong, “Dynamic and multi- faceted spatio-temporal deep learning for traffic speed forecasting,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2021, pp. 547–555

    2021

  26. [35]

    Spatial-temporal attention network for crime prediction with adaptive graph learning,

    M. Sun, P. Zhou, H. Tian, Y . Liao, and H. Xie, “Spatial-temporal attention network for crime prediction with adaptive graph learning,” in ICANN. Springer, 2022, pp. 656–669

    2022

  27. [36]

    Spatiotemporal adaptive gated graph convolution network for urban traffic flow forecasting,

    B. Lu, X. Gan, H. Jin, L. Fu, and H. Zhang, “Spatiotemporal adaptive gated graph convolution network for urban traffic flow forecasting,” in CIKM, 2020, pp. 1025–1034

    2020

  28. [37]

    Adapgl: An adaptive graph learning algorithm for traffic prediction based on spatiotemporal neural networks,

    W. Zhang, F. Zhu, Y . Lv, C. Tan, W. Liu, X. Zhang, and F.-Y . Wang, “Adapgl: An adaptive graph learning algorithm for traffic prediction based on spatiotemporal neural networks,” TRC, vol. 139, p. 103659, 2022. JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, AUGUST 201X 12

    2022

  29. [38]

    Freeway performance mea- surement system: operational analysis tool,

    T. Choe, A. Skabardonis, and P. Varaiya, “Freeway performance mea- surement system: operational analysis tool,” Transportation research record, vol. 1811, no. 1, pp. 67–75, 2002

    2002

  30. [39]

    A comprehensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and S. Y . Philip, “A comprehensive survey on graph neural networks,” IEEE transactions on neural networks and learning systems , vol. 32, no. 1, pp. 4–24, 2020

    2020

  31. [40]

    Unist: a prompt-empowered universal model for urban spatio-temporal prediction,

    Y . Yuan, J. Ding, J. Feng, D. Jin, and Y . Li, “Unist: a prompt-empowered universal model for urban spatio-temporal prediction,” in Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024, pp. 4095–4106

    2024

  32. [41]

    Evaluating the generalization ability of spatiotemporal model in urban scenario,

    H. Wang, J. Chen, T. Pan, Z. Dong, L. Zhang, R. Jiang, and X. Song, “Evaluating the generalization ability of spatiotemporal model in urban scenario,” arXiv preprint arXiv:2410.04740 , 2024

    2024 arXiv

  33. [42]

    Urban traffic prediction from spatio-temporal data using deep meta learning,

    Z. Pan, Y . Liang, W. Wang, Y . Yu, Y . Zheng, and J. Zhang, “Urban traffic prediction from spatio-temporal data using deep meta learning,” in Proceedings of the 25th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2019, pp. 1720–1730

    2019

  34. [43]

    Cross-city transfer learning for deep spatio-temporal prediction,

    L. Wang, X. Geng, X. Ma, F. Liu, and Q. Yang, “Cross-city transfer learning for deep spatio-temporal prediction,” arXiv preprint arXiv:1802.00386, 2018

    2018 arXiv

  35. [44]

    Analysis of microarray data using z score transformation,

    C. Cheadle, M. P. Vawter, W. J. Freed, and K. G. Becker, “Analysis of microarray data using z score transformation,” The Journal of molecular diagnostics, vol. 5, no. 2, pp. 73–81, 2003

    2003

  36. [45]

    Dynamic graph convolution network with spatio-temporal attention fusion for traffic flow prediction,

    X. Luo, C. Zhu, D. Zhang, and Q. Li, “Dynamic graph convolution network with spatio-temporal attention fusion for traffic flow prediction,” arXiv preprint arXiv:2302.12598 , 2023

    2023 arXiv

  37. [46]

    Con- necting the dots: Multivariate time series forecasting with graph neural networks,

    Z. Wu, S. Pan, G. Long, J. Jiang, X. Chang, and C. Zhang, “Con- necting the dots: Multivariate time series forecasting with graph neural networks,” in Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , 2020, pp. 753–763

    2020

  38. [47]

    Coupled layer-wise graph convolution for transportation demand prediction,

    J. Ye, L. Sun, B. Du, Y . Fu, and H. Xiong, “Coupled layer-wise graph convolution for transportation demand prediction,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 35, no. 5, 2021, pp. 4617–4625

    2021

  39. [48]

    Dynamic and multi- faceted spatio-temporal deep learning for traffic speed forecasting,

    L. Han, B. Du, L. Sun, Y . Fu, Y . Lv, and H. Xiong, “Dynamic and multi- faceted spatio-temporal deep learning for traffic speed forecasting,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining , 2021, pp. 547–555

    2021

  40. [49]

    Dynamic graph convolutional recurrent network for traffic prediction: Benchmark and solution,

    F. Li, J. Feng, H. Yan, G. Jin, F. Yang, F. Sun, D. Jin, and Y . Li, “Dynamic graph convolutional recurrent network for traffic prediction: Benchmark and solution,” ACM Transactions on Knowledge Discovery from Data, pp. 1–21, 2021

    2021

  41. [50]

    Spatio-temporal diffusion point processes,

    Y . Yuan, J. Ding, C. Shao, D. Jin, and Y . Li, “Spatio-temporal diffusion point processes,” in Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023

    2023

  42. [51]

    Robust spatiotemporal traffic forecasting with reinforced dynamic adversarial training,

    F. Liu, W. Zhang, and H. Liu, “Robust spatiotemporal traffic forecasting with reinforced dynamic adversarial training,” in Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023

    2023

  43. [52]

    Greto: rem- edying dynamic graph topology-task discordance via target homophily,

    Z. Zhou, Q. Huang, G. Lin, K. Yang, L. Bai, and Y . Wang, “Greto: rem- edying dynamic graph topology-task discordance via target homophily,” in The eleventh international conference on learning representations , 2022

    2022

  44. [53]

    Predicting collective human mobility via countering spatiotemporal heterogeneity,

    Z. Zhou, K. Yang, Y . Liang, B. Wang, H. Chen, and Y . Wang, “Predicting collective human mobility via countering spatiotemporal heterogeneity,” IEEE Transactions on Mobile Computing , 2023

    2023

  45. [54]

    Learning gaussian mixture representations for tensor time series forecasting,

    J. Deng, J. Deng, R. Jiang, and X. Song, “Learning gaussian mixture representations for tensor time series forecasting,” in Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence , 2023

    2023

  46. [55]

    Multi-modality spatio- temporal forecasting via self-supervised learning,

    J. Deng, R. Jiang, J. Zhang, and X. Song, “Multi-modality spatio- temporal forecasting via self-supervised learning,” arXiv preprint arXiv:2405.03255, 2024

    2024 arXiv

  47. [56]

    Utilizing real-world transporta- tion data for accurate traffic prediction,

    B. Pan, U. Demiryurek, and C. Shahabi, “Utilizing real-world transporta- tion data for accurate traffic prediction,” in 2012 ieee 12th international conference on data mining . IEEE, 2012, pp. 595–604

    2012

  48. [57]

    Vector autoregressions,

    J. H. Stock and M. W. Watson, “Vector autoregressions,” Journal of Economic perspectives, vol. 15, no. 4, pp. 101–115, 2001

    2001

  49. [58]

    Nonparametric regression and short-term freeway traffic forecasting,

    G. A. Davis and N. L. Nihan, “Nonparametric regression and short-term freeway traffic forecasting,” Journal of Transportation Engineering, vol. 117, no. 2, pp. 178–188, 1991

    1991

  50. [59]

    A comparison of the performance of artificial neural networks and support vector machines for the prediction of traffic speed,

    L. Vanajakshi and L. R. Rilett, “A comparison of the performance of artificial neural networks and support vector machines for the prediction of traffic speed,” in IEEE Intelligent Vehicles Symposium, 2004. IEEE, 2004, pp. 194–199

    2004

  51. [60]

    Spectral temporal graph neural network for multivariate time-series forecasting,

    D. Cao, Y . Wang, J. Duan, C. Zhang, X. Zhu, C. Huang, Y . Tong, B. Xu, J. Bai, J. Tong et al. , “Spectral temporal graph neural network for multivariate time-series forecasting,” Advances in neural information processing systems, vol. 33, pp. 17 766–17 778, 2020

    2020

  52. [61]

    A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection,

    M. Jin, H. Y . Koh, Q. Wen, D. Zambon, C. Alippi, G. I. Webb, I. King, and S. Pan, “A survey on graph neural networks for time series: Forecasting, classification, imputation, and anomaly detection,” arXiv preprint arXiv:2307.03759, 2023

    2023 arXiv

  53. [62]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” arXiv preprint arXiv:1609.02907 , 2016

    2016 arXiv

  54. [63]

    Graph attention networks,

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Lio, and Y . Bengio, “Graph attention networks,” 2018

    2018

  55. [64]

    Inductive representation learning on large graphs,

    W. Hamilton, Z. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” Advances in neural information processing systems, vol. 30, 2017

    2017

  56. [65]

    Adaptive universal generalized pagerank graph neural network,

    E. Chien, J. Peng, P. Li, and O. Milenkovic, “Adaptive universal generalized pagerank graph neural network,” in ICLR, 2021. [Online]. Available: https://openreview.net/forum?id=n6jl7fLxrP

    2021

  57. [66]

    Predict then propagate: Graph neural networks meet personalized pagerank,

    J. Klicpera, A. Bojchevski, and S. G ¨unnemann, “Predict then propagate: Graph neural networks meet personalized pagerank,” arXiv preprint arXiv:1810.05997, 2018

    2018 arXiv

  58. [67]

    A comprehen- sive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, “A comprehen- sive survey on graph neural networks,”arXiv preprint arXiv:1901.00596, 2019

    1901 arXiv

  59. [68]

    Shift-robust gnns: Overcoming the limitations of localized graph training data,

    Q. Zhu, N. Ponomareva, J. Han, and B. Perozzi, “Shift-robust gnns: Overcoming the limitations of localized graph training data,” Advances in Neural Information Processing Systems , vol. 34, 2021

    2021

  60. [69]

    Handling distribution shifts on graphs: An invariance perspective,

    Q. Wu, H. Zhang, J. Yan, and D. Wipf, “Handling distribution shifts on graphs: An invariance perspective,” in International Conference on Learning Representations , 2022. [Online]. Available: https://openreview.net/forum?id=FQOC5u-1egI

    2022

  61. [70]

    Confidence may cheat: Self-training on graph neural networks under distribution shift,

    H. Liu, B. Hu, X. Wang, C. Shi, Z. Zhang, and J. Zhou, “Confidence may cheat: Self-training on graph neural networks under distribution shift,” in Proceedings of the ACM Web Conference 2022, 2022, pp. 1248–1258

    2022

  62. [71]

    Invariance principle meets out-of-distribution generalization on graphs,

    Y . Chen, Y . Zhang, H. Yang, K. Ma, B. Xie, T. Liu, B. Han, and J. Cheng, “Invariance principle meets out-of-distribution generalization on graphs,” arXiv preprint arXiv:2202.05441 , 2022

    2022 arXiv

  63. [72]

    Sizeshiftreg: a regularization method for improving size-generalization in graph neural networks,

    D. Buffelli, P. Li `o, and F. Vandin, “Sizeshiftreg: a regularization method for improving size-generalization in graph neural networks,” Advances in Neural Information Processing Systems , 2022

    2022

  64. [73]

    Good: A graph out-of-distribution benchmark,

    S. Gui, X. Li, L. Wang, and S. Ji, “Good: A graph out-of-distribution benchmark,” arXiv preprint arXiv:2206.08452 , 2022

    2022 arXiv

  65. [74]

    Discovering invariant rationales for graph neural networks,

    Y . Wu, X. Wang, A. Zhang, X. He, and T.-S. Chua, “Discovering invariant rationales for graph neural networks,” in International Conference on Learning Representations , 2022. [Online]. Available: https://openreview.net/forum?id=hGXij5rfiHw

    2022

  66. [75]

    Graph domain adaptation via theory-grounded spectral regularization,

    Y . You, T. Chen, Z. Wang, and Y . Shen, “Graph domain adaptation via theory-grounded spectral regularization,” in International Conference on Learning Representations, 2023. Hongjun Wang is working toward the PhD degree in the Department of Mechano-Informatics at The Universi...

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.