Pith. sign in

REVIEW 4 major objections 4 minor 52 references

Channel-Independent Federated Traffic Prediction

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Federated traffic prediction can be done without inter-client data communication, and the paper claims this is not only cheaper but also more accurate.

desk verdict A sensible adaptation of channel independence to federated traffic prediction, but the headline outperformance claim is contradicted by the paper's own Table 3. read the letter →

arxiv 2508.04517 v1 pith:QRTED22F submitted 2025-08-06 cs.LG

Mo Zhang , Xiaoyu Li , Bin Xu , Meng Chen , Yongshun Gong This is my paper
classification cs.LG
keywords federatedlearningtrafficpredictionchannelindependencecommunicationefficiencyspatiotemporalforecastingMLPnodeembeddingFedAvg
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

This paper argues that the standard assumption in federated traffic prediction—that clients must exchange node data or model parts to capture spatial dependencies—is unnecessary. It proposes the Channel-Independent Paradigm (CIP), under which each sensor node forecasts future traffic from its own history alone, and Fed-CI, a federated MLP framework built on it. The paper claims Fed-CI matches or beats graph-based federated baselines on four real-world traffic datasets while reducing communication cost to just model-parameter aggregation, cutting training time sharply, and improving RMSE, MAE, and MAPE by roughly 8%, 14%, and 16%.

What carries the argument

The Channel-Independent Paradigm (CIP), which treats each sensor node as an independent channel whose future depends only on its own past, carries the argument. Fed-CI implements it with an MLP backbone that combines one-hot time-of-day/week embeddings, node-embedding codebooks, a personalized client bias, and a temporal MLP block. The federated aggregation rule, FedEmbedAvg, averages the shared parameters across clients while updating node-embedding rows only from the client that owns the corresponding nodes, keeping parameter shapes consistent and enabling the zero-data-exchange design.

What would settle it

Take a traffic network with a known bottleneck where upstream congestion strongly determines downstream flow, split it across clients, and train Fed-CI. If its error on downstream sensors exceeds that of a channel-dependent federated model (e.g., CNFGNN) by a meaningful margin, the channel-independence premise fails in exactly the spatial-propagation regime where federated graph methods are supposed to help. Alternatively, compare Fed-CI against a centralized model trained on all nodes' data: if Fed-CI's error approaches the centralized error, the missing spatial information is immaterial; if

Watch

Extended reading notes

Core claim

The discovery is that channel independence, previously shown to be competitive in centralized multivariate time-series forecasting, transfers to federated traffic prediction: a per-node, per-client MLP with learnable time and node embeddings plus a client-specific bias can predict traffic accurately without any cross-node data sharing. Fed-CI eliminates the data-communication term from the cost equation, leaving only the usual FedAvg-style parameter exchange, and achieves state-of-the-art results on PeMSD4, PeMSD8, PEMS-BAY, and METR-LA across several federated settings. The paper presents this as breaking the trade-off between communication cost and prediction accuracy that current federate

Load-bearing premise

The load-bearing premise is that each sensor's future traffic can be predicted as accurately from its own history alone as from the entire road network's history; this is imported from centralized time-series work and is not validated on the federated traffic datasets used here.

Editorial extensions

If this is right

  • Communication cost in federated traffic prediction drops to the model-aggregation term, since the data-exchange term becomes zero.
  • Training time per global epoch falls sharply, making large-scale federated traffic learning feasible on commodity networks; the paper reports a drop from 23820 s to 313 s per global epoch on PEMS-BAY with 325 clients.
  • Lightweight MLP clients can run on resource-constrained edge devices, since no graph convolution or server-side spatial model is required.
  • Each sensor's prediction is attributable to its own past plus time and node identity, which simplifies debugging and explanation relative to graph-based models.
  • Privacy is strengthened because raw node data never leaves the client; only parameter updates reach the server.

Reading between the lines

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

  • If CIP holds for road sensors, the same 'drop the cross-client channel' design is likely to transfer to other federated spatiotemporal tasks such as energy load, crowd flow, or air-quality forecasting, where spatial correlation is similarly assumed to be load-bearing.
  • Fed-CI's accuracy is measured against federated baselines, not against a centralized oracle with access to all nodes; the paper therefore does not quantify how much accuracy is sacrificed for privacy, which would bound the real cost of channel independence.
  • A stress test on a network with strong, time-lagged spatial propagation—say, a highway bottleneck where upstream congestion determines downstream flow—would define the regime where CIP fails; if downstream prediction error grows when upstream data is unavailable, a hybrid that exchanges only task-critical spatial messages may be needed.
  • FedEmbedAvg's row-wise node-embedding update could be composed with secure aggregation or differential-privacy mechanisms to provide formal privacy guarantees beyond the structural separation of data.
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 / 4 minor

Summary. The paper proposes a channel-independent paradigm (CIP) for federated traffic prediction, implemented as Fed-CI. Instead of exchanging node-level data or model parts across clients to model inter-node dependencies, Fed-CI lets each node predict from its own history using a lightweight MLP with time and node embeddings plus a personalized client bias. The server aggregates shared parameters with a FedEmbedAvg strategy that averages all parameters except node embeddings, whose rows are updated from the owning client. The authors claim that Fed-CI eliminates inter-client data communication, reduces communication cost and training time, and achieves state-of-the-art prediction accuracy, with average improvements of 8.04%, 14.35%, and 16.33% in RMSE, MAE, and MAPE. Experiments are reported on PeMSD4, PeMSD8, PEMS-BAY, and METR-LA under several federated settings, with comparisons to HI, MLP, GRU, LSTM, and, in some settings, MFVSTGNN, FLoS, FCGCN, CNFGNN, FedGTP, and CTFL.

Significance. If the empirical claims were supported, the core idea would be valuable: showing that a zero-data-exchange, channel-independent MLP can be competitive in federated traffic prediction would substantially reduce communication overhead and simplify deployment. The communication-cost analysis in Table 1 and the multi-GPU socket-based simulation are concrete strengths, and the ablation in Table 5 demonstrates that the time/node embeddings matter. However, the central claim of consistent state-of-the-art performance is not supported by the paper's own experimental table, and the comparison set is incomplete by the authors' own admission. The significance of the contribution therefore hinges on exactly the evidence that is weakest.

major comments (4)
  1. [Abstract; §6.4; Table 3] The abstract and §6.4 claim that Fed-CI 'consistently outperforms existing methods across all datasets and federated settings.' Table 3 contradicts this. In METR-LA (8 clients, 12→12), Fed-CI has RMSE 5.88, while FedGTP reports 4.41 and MFVSTGNN reports 4.45; in PeMSD4 (28 clients, 6→1) occupancy, FedGTP has RMSE 0.0126 versus Fed-CI 0.0130. Thus the 'consistently outperforms' and 'state-of-the-art' statements are false as written. The averaged improvement figures (8.04%, 14.35%, 16.33%) are not derived from any stated formula or set of rows in Table 3.
  2. [Table 3 caption; §6.3] The table caption states that many advanced baselines are marked '–' because the original papers do not report those settings and because the authors' reproduction attempts failed to match reported performance, so the comparison was abandoned. This is a serious selection issue for a state-of-the-art claim: MFVSTGNN, FLoS, FCGCN, and CNFGNN are absent from most rows, and strong competitors appear only where they are favorable to the narrative. To support a SOTA claim the authors need to include the original reported numbers as a separate row/column, or systematically explain and justify every exclusion, rather than dropping non-reproducing baselines from the comparison.
  3. [§6.2; Table 3] No error bars, confidence intervals, or number of random seeds are reported. Several comparisons rest on small margins, e.g., PeMSD8 speed MAPE 1.23 vs. 1.35 and PeMSD4 occupancy MAE 0.0057 vs. 0.0064. Without repeated runs or a statistical test, the claimed superiority over FedGTP and other baselines is not established even in the rows where Fed-CI is numerically best.
  4. [§4.2] The channel-independence premise is imported from prior centralized multivariate time-series work ([15]) and is not validated on the federated traffic datasets used here. If spatial propagation matters for these sensors, the zero-data-communication design could fail exactly where graph-based federated methods are intended to help. A concrete test would be a complete comparison against a channel-dependent federated model under identical client partitions, or a centralized upper-bound analysis. The current table is too incomplete to serve as that test.
minor comments (4)
  1. [§5.1, Eqs. (4)–(7)] The notation for one-hot indices is confusing: T_d and T_w are described as index tensors but written with an extra dimension N_d/N_w, and the sentence 'one-hot encode these indices using the one-hot indices' is awkward. Please clarify the shapes and the one-hot operation.
  2. [§5.4, Eqs. (11)–(13)] The same symbol E is used for the tensor before and after the transpose, which makes Eq. (11) hard to parse. Use E and E' consistently.
  3. [Abstract; §1; §6.4] The reported improvements are inconsistent: the abstract rounds to 8%, 14%, 16%, the introduction gives 8.04%, 14.35%, 16.33%, and §6.4 says 'most settings' rather than 'all settings.' These numbers and the strength of the wording should be aligned.
  4. [§6.2] The data split is described only as 'depend[ing] on the specific federated learning setup, following previous work.' The exact train/validation/test split for each setting should be stated explicitly for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Fed-CI's accuracy claims are empirical comparisons against external baselines; the channel-independence premise is imported from external prior work, not derived from the paper's own fitted quantities.

full rationale

The paper's central claim is empirical: Fed-CI outperforms baselines on real-world traffic datasets. These results are measured against external models (CNFGNN, FedGTP, MFVSTGNN, etc.) and standard datasets, not derived from a parameter fitted to the same target values. The channel-independence premise is imported from an external prior study [15] and from ST-MLP [33], neither of which is authored by the present paper's authors, so there is no self-citation chain that forces the conclusion. The communication-cost reduction is a design property of the channel-independent paradigm rather than a fitted empirical prediction, and it is separately quantified. The paper does cite the authors' own prior works ([3], [19]), but these appear only as peripheral related-work mentions for transformer/GCN approaches and are not load-bearing for the Fed-CI derivation. There is no uniqueness theorem imported from the authors, no ansatz smuggled in via self-citation, and no renamed known result presented as a derivation. The abstract's 'consistently outperforms' claim is contradicted by the paper's own Table 3 on METR-LA (8 clients, 12→12), where FedGTP and MFVSTGNN achieve lower RMSE than Fed-CI, and several baselines are missing from many settings; however, this is a correctness/comparison issue, not circularity. The central derivation chain is self-contained: the model architecture is described by explicit equations, and the performance claims are externally falsifiable. Therefore no circular step is present.

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

The central claim rests on a domain assumption (channel independence) imported from centralized time series work, plus standard federated learning assumptions about node disjointness and FedAvg validity. The free parameters are ordinary hyperparameters and client partition choices, not exotic fitting constants. No new physical or conceptual entities are introduced.

free parameters (3)
  • number of MLP layers = 3 (12 was best in sensitivity analysis)
    Section 6.6.3: the authors choose 3 layers to reduce model complexity, despite 12 layers giving the best RMSE/MAE/MAPE.
  • embedding dimensions (d_Td, d_Tw, d_n) = Not explicitly reported; chosen via sensitivity analysis over powers of two
    Sections 5.1 and 6.6.3: dimensions are varied from 2 to 256 and the final values are not stated in the text.
  • client partition and number of clients per setting = 325, 207, 8, 5, 28, 14 clients depending on setting
    Section 6 and Table 3: the federated settings are chosen manually; node assignment to clients is not described.
assumptions (3)
  • domain assumption Channel independence is a valid inductive bias for traffic prediction: each node's future depends only on its own history.
    Section 4.2 relies on this to justify zero data communication; it is imported from prior work [15] and ST-MLP [33], not validated for the federated traffic datasets used.
  • domain assumption Clients hold non-overlapping node sets, so node embeddings can be aggregated by direct row overwrite (FedEmbedAvg).
    Algorithm 2 assumes each client's nodes are disjoint; this is stated in Figure 1 but the node assignment process is not experimentally controlled or described.
  • domain assumption FedAvg over the non-embedding parameters is a valid aggregation under client data heterogeneity.
    Algorithm 2 averages remaining parameters across clients; standard federated learning assumptions, but the paper does not analyze non-IID effects in this setting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Channel-Independent Federated Traffic Prediction." pith.science (2026). https://pith.science/paper/QRTED22F

@misc{pith2026250804517,
  author       = {Pith},
  title        = {Pith review of: Channel-Independent Federated Traffic Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QRTED22F}},
  note         = {Machine review of arXiv:2508.04517}
}
read the original abstract

In recent years, traffic prediction has achieved remarkable success and has become an integral component of intelligent transportation systems. However, traffic data is typically distributed among multiple data owners, and privacy constraints prevent the direct utilization of these isolated datasets for traffic prediction. Most existing federated traffic prediction methods focus on designing communication mechanisms that allow models to leverage information from other clients in order to improve prediction accuracy. Unfortunately, such approaches often incur substantial communication overhead, and the resulting transmission delays significantly slow down the training process. As the volume of traffic data continues to grow, this issue becomes increasingly critical, making the resource consumption of current methods unsustainable. To address this challenge, we propose a novel variable relationship modeling paradigm for federated traffic prediction, termed the Channel-Independent Paradigm(CIP). Unlike traditional approaches, CIP eliminates the need for inter-client communication by enabling each node to perform efficient and accurate predictions using only local information. Based on the CIP, we further develop Fed-CI, an efficient federated learning framework, allowing each client to process its own data independently while effectively mitigating the information loss caused by the lack of direct data sharing among clients. Fed-CI significantly reduces communication overhead, accelerates the training process, and achieves state-of-the-art performance while complying with privacy regulations. Extensive experiments on multiple real-world datasets demonstrate that Fed-CI consistently outperforms existing methods across all datasets and federated settings. It achieves improvements of 8%, 14%, and 16% in RMSE, MAE, and MAPE, respectively, while also substantially reducing communication costs.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 47 canonical work pages

  1. [15]

    Lu Han, Han-Jia Ye, and De-Chuan Zhan. 2024. The capacity and robustness trade-off: Revisiting the channel independent strategy for multivariate time series forecasting. IEEE Transactions on Knowledge and Data Engineering (2024)

    work page 2024

  2. [1]

    Sawsan Abdulrahman, Hanine Tout, Hakima Ould-Slimane, Azzam Mourad, Chamseddine Talhi, and Mohsen Guizani. 2021. A Survey on Federated Learning: The Journey From Centralized to Distributed On-Site Learning and Beyond.IEEE Internet of Things Journal 8, 7 (2021), 5476–5497. doi:10.1109/JIOT.2020.3030072

    work page arXiv 2021

  3. [2]

    Muhammad Alam, Joaquim Ferreira, and José Fonseca. 2016. Introduction to Intelligent Transportation Systems . Springer International Publishing, Cham, 1–17. doi:10.1007/978-3-319-28183-4_1

    work page doi:10.1007/978-3-319-28183-4_1 2016

  4. [3]

    Yang An, Zhibin Li, Wei Liu, Haoliang Sun, Meng Chen, Wenpeng Lu, and Yongshun Gong. 2024. Spatio-temporal Graph Normalizing Flow for Probabilistic Traffic Prediction. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management . 45–55

    work page 2024

  5. [4]

    Lei Bai, Lina Yao, Can Li, Xianzhi Wang, and Can Wang. 2020. Adaptive graph convolutional recurrent network for traffic forecasting. Advances in Neural Information Processing Systems 33 (2020), 17804–17815

    work page 2020

  6. [5]

    Ling Cai, Krzysztof Janowicz, Gengchen Mai, Bo Yan, and Rui Zhu. 2020. Traffic transformer: Capturing the continuity and periodicity of time series for traffic forecasting. Transactions in GIS 24, 3 (2020), 736–755

    work page 2020

  7. [6]

    Changlu Chen, Yanbin Liu, Ling Chen, and Chengqi Zhang. 2023. Bidirectional Spatial-Temporal Adaptive Transformer for Urban Traffic Flow Forecasting.IEEE Transactions on Neural Networks and Learning System 34, 10 (2023), 6913–6925

    work page 2023

  8. [7]

    Weiqi Chen, Ling Chen, Yu Xie, Wei Cao, Yusong Gao, and Xiaojie Feng. 2020. Multi-range attentive bicomponent graph convolutional network for traffic fore- casting. In Proceedings of the AAAI conference on artificial intelligence , Vol. 34. 3529–3536

    work page 2020

Show all 52 references
  1. [8]

    Yue Cui, Jiandong Xie, and Kai Zheng. 2021. Historical Inertia: A Neglected but Powerful Baseline for Long Sequence Time-series Forecasting. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management (Virtual Event, Queensland, Australia) (CI...

    2021

  2. [9]

    George Dimitrakopoulos and Panagiotis Demestichas. 2010. Intelligent Trans- portation Systems. IEEE Vehicular Technology Magazine 5, 1 (2010), 77–84. doi:10.1109/MVT.2009.935537

    2010

  3. [10]

    Vijay Ekambaram, Arindam Jati, Nam Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. 2023. Tsmixer: Lightweight mlp-mixer model for multivariate time series forecasting. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 459–469

    2023

  4. [11]

    Aosong Feng and Leandros Tassiulas. 2022. Adaptive graph spatial-temporal transformer network for traffic forecasting. In Proceedings of the 31st ACM inter- national conference on information & knowledge management . 3933–3937

    2022

  5. [12]

    Xu Geng, Yaguang Li, Leye Wang, Lingyu Zhang, Qiang Yang, Jieping Ye, and Yan Liu. 2019. Spatiotemporal multi-graph convolution network for ride-hailing demand forecasting. InProceedings of the AAAI conference on artificial intelligence, Vol. 33. 3656–3663

    2019

  6. [13]

    Kan Guo, Yongli Hu, Zhen Qian, Yanfeng Sun, Junbin Gao, and Baocai Yin

  7. [14]

    Shengnan Guo, Youfang Lin, Ning Feng, Chao Song, and Huaiyu Wan. 2019. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. In Proceedings of the AAAI conference on artificial intelligence , Vol. 33. 922–929

    2019

  8. [16]

    Jiawei Jiang, Chengkai Han, Wayne Xin Zhao, and Jingyuan Wang. 2023. PDFormer: Propagation Delay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction. In AAAI Conference on Artificial Intelligence . 4365–4373

    2023

  9. [17]

    Shiyong Lan, Yitong Ma, Weikang Huang, Wenwu Wang, Hongyu Yang, and Pyang Li. 2022. Dstagnn: Dynamic spatial-temporal aware graph neural network for traffic flow forecasting. In International Conference on Machine Learning . PMLR, 11906–11917

    2022

  10. [18]

    Fuxian Li, Huan Yan, Guangyin Jin, Yue Liu, Yong Li, and Depeng Jin. 2022. Automated spatio-temporal synchronous modeling with multiple graphs for traffic prediction. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management . 1084–1093

    2022

  11. [19]

    Xiaoyu Li, Yongshun Gong, Wei Liu, Yilong Yin, Yu Zheng, and Liqiang Nie. 2024. Dual-track Spatio-temporal Learning for Urban Flow Prediction with Adaptive Normalization. Artificial Intelligence (2024), 104065

    2024

  12. [20]

    Yaguang Li, Rose Yu, Cyrus Shahabi, and Yan Liu. 2018. Diffusion Convolu- tional Recurrent Neural Network: Data-Driven Traffic Forecasting. In Interna- tional Conference on Learning Representations . https://openreview.net/forum? id=SJiHXGWAZ

    2018

  13. [21]

    Hangchen Liu, Zheng Dong, Renhe Jiang, Jiewen Deng, Jinliang Deng, Quanjun Chen, and Xuan Song. 2023. Spatio-temporal adaptive embedding makes vanilla transformer sota for traffic forecasting. InInternational Conference on Information and Knowledge Management. 4125–4129

    2023

  14. [22]

    Lei Liu, Yuxing Tian, Chinmay Chakraborty, Jie Feng, Qingqi Pei, Li Zhen, and Keping Yu. 2023. Multilevel federated learning-based intelligent traffic flow fore- casting for transportation network management. IEEE Transactions on Network and Service Management 20, 2 (2023), 1446–1458

    2023

  15. [23]

    Lingbo Liu, Jiajie Zhen, Guanbin Li, Geng Zhan, Zhaocheng He, Bowen Du, and Liang Lin. 2020. Dynamic spatial-temporal representation learning for traffic flow prediction. IEEE Transactions on Intelligent Transportation Systems 22, 11 (2020), 7169–7183

    2020

  16. [24]

    Xu Liu, Yuxuan Liang, Chao Huang, Hengchang Hu, Yushi Cao, Bryan Hooi, and Roger Zimmermann. 2023. Do We Really Need Graph Neural Networks for Traffic Forecasting? CoRR abs/2301.12603 (2023)

    2023 arXiv

  17. [25]

    Yi Liu, JQ James, Jiawen Kang, Dusit Niyato, and Shuyu Zhang. 2020. Privacy- preserving traffic flow prediction: A federated learning approach. IEEE Internet of Things Journal 7, 8 (2020), 7751–7763

    2020

  18. [26]

    Chuizheng Meng, Sirisha Rambhatla, and Yan Liu. 2021. Cross-node federated graph neural network for spatio-temporal data modeling. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining . 1202–1211

    2021

  19. [27]

    Tao Qi, Lingqiang Chen, Guanghui Li, Yijing Li, and Chenshu Wang. 2023. FedAGCN: A traffic flow prediction framework based on federated learning and Asynchronous Graph Convolutional Network. Applied Soft Computing 138 (2023), 110175

    2023

  20. [28]

    Zezhi Shao, Fei Wang, Yongjun Xu, Wei Wei, Chengqing Yu, Zhao Zhang, Di Yao, Tao Sun, Guangyin Jin, Xin Cao, et al. 2024. Exploring progress in multi- variate time series forecasting: Comprehensive benchmarking and heterogeneity analysis. IEEE Transactions on Knowledge and Dat...

    2024

  21. [29]

    Zezhi Shao, Zhao Zhang, Fei Wang, Wei Wei, and Yongjun Xu. 2022. Spatial- temporal identity: A simple yet effective baseline for multivariate time series forecasting. InProceedings of the 31st ACM International Conference on Information & Knowledge Management. 4454–4458

    2022

  22. [30]

    Chao Song, Youfang Lin, Shengnan Guo, and Huaiyu Wan. 2020. Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial- Temporal Network Data Forecasting. InAAAI Conference on Artificial Intelligence. 914–921

    2020

  23. [31]

    Ilya O Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Andreas Steiner, Daniel Keysers, Jakob Uszkoreit, et al. 2021. Mlp-mixer: An all-mlp architecture for vision. Advances in neural information processing systems 3...

    2021

  24. [32]

    Hanqiu Wang, Rongqing Zhang, Xiang Cheng, and Liuqing Yang. 2022. Federated Spatio-Temporal Traffic Flow Prediction Based on Graph Convolutional Network. In 2022 14th International Conference on Wireless Communications and Signal Processing (WCSP). IEEE, 221–225

    2022

  25. [33]

    Zepu Wang, Yuqi Nie, Peng Sun, Nam H Nguyen, John Mulvey, and H Vincent Poor. 2023. St-mlp: A cascaded spatio-temporal linear framework with channel- independence strategy for traffic forecasting. arXiv preprint arXiv:2308.07496 (2023)

    2023 arXiv

  26. [34]

    Gerald Woo, Chenghao Liu, Doyen Sahoo, Akshat Kumar, and Steven Hoi. 2023. Learning deep time-index models for time series forecasting. In International Conference on Machine Learning . PMLR, 37217–37237

    2023

  27. [35]

    Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, Xiaojun Chang, and Chengqi Zhang. 2020. Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks. In ACM SIGKDD International Conference on Knowledge Discovery & Data Mining . 753–763

    2020

  28. [36]

    Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, and Chengqi Zhang. 2019. Graph WaveNet for Deep Spatial-Temporal Graph Modeling. InProceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, IJCAI-19 . International Joint Conferences on Artifi...

    2019

  29. [37]

    Mengran Xia, Dawei Jin, and Jingyu Chen. 2022. Short-term traffic flow prediction based on graph convolutional networks and federated learning.IEEE Transactions on Intelligent Transportation Systems 24, 1 (2022), 1191–1203

    2022

  30. [38]

    Zhijian Xu, Ailing Zeng, and Qiang Xu. 2024. FITS: Modeling Time Series with $10k$ Parameters. In The Twelfth International Conference on Learning Representations. https://openreview.net/forum?id=bWcnvZ3qMb

    2024

  31. [39]

    Linghua Yang, Wantong Chen, Xiaoxi He, Shuyue Wei, Yi Xu, Zimu Zhou, and Yongxin Tong. 2024. FedGTP: Exploiting inter-client spatial dependency in federated graph-based traffic prediction. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining ...

    2024

  32. [40]

    Kun Yi, Qi Zhang, Wei Fan, Hui He, Liang Hu, Pengyang Wang, Ning An, Long- bing Cao, and Zhendong Niu. 2024. FourierGNN: Rethinking multivariate time Channel-Independent Federated Traffic Prediction series forecasting from a pure graph perspective. Advances in Neural Informati...

    2024

  33. [41]

    Xueyan Yin, Genze Wu, Jinze Wei, Yanming Shen, Heng Qi, and Baocai Yin. 2022. Deep Learning on Traffic Prediction: Methods, Analysis, and Future Directions. IEEE Transactions on Intelligent Transportation Systems 23, 6 (2022), 4927–4943. doi:10.1109/TITS.2021.3054840

    2022

  34. [42]

    Bing Yu, Haoteng Yin, and Zhanxing Zhu. 2018. Spatio-Temporal Graph Con- volutional Networks: A Deep Learning Framework for Traffic Forecasting. In Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI-18. International Joint Confer...

    2018 doi

  35. [43]

    Haitao Yuan and Guoliang Li. 2021. A survey of traffic prediction: from spatio- temporal data to intelligent transportation. Data Science and Engineering 6, 1 (2021), 63–85

    2021

  36. [44]

    Ailing Zeng, Muxi Chen, Lei Zhang, and Qiang Xu. 2023. Are transformers effective for time series forecasting?. In Proceedings of the AAAI conference on artificial intelligence, Vol. 37. 11121–11128

    2023

  37. [45]

    Chen Zhang, Yu Xie, Hang Bai, Bin Yu, Weihong Li, and Yuan Gao. 2021. A survey on federated learning. Knowledge-Based Systems 216 (2021), 106775

    2021

  38. [46]

    Chenhan Zhang, Shuyu Zhang, JQ James, and Shui Yu. 2021. FASTGNN: A topological information protected federated learning approach for traffic speed forecasting. IEEE Transactions on Industrial Informatics 17, 12 (2021), 8464–8474

    2021

  39. [47]

    Chenhan Zhang, Shiyao Zhang, Shui Yu, and JQ James. 2022. Graph-based traffic forecasting via communication-efficient federated learning. In2022 IEEE Wireless Communications and Networking Conference (WCNC) . IEEE, 2041–2046

    2022

  40. [48]

    Zijian Zhang, Ze Huang, Zhiwei Hu, Xiangyu Zhao, Wanyu Wang, Zitao Liu, Junbo Zhang, S Joe Qin, and Hongwei Zhao. 2023. MLPST: MLP is All You Need for Spatio-Temporal Prediction. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management . ...

    2023

  41. [49]

    Liang Zhao, Min Gao, and Zongwei Wang. 2022. St-gsp: Spatial-temporal global semantic representation learning for urban flow prediction. In International Conference on Web Search and Data Mining . 1443–1451

    2022

  42. [50]

    Chuanpan Zheng, Xiaoliang Fan, Cheng Wang, and Jianzhong Qi. 2020. Gman: A graph multi-attention network for traffic prediction. In Proceedings of the AAAI conference on artificial intelligence , Vol. 34. 1234–1241

    2020

  43. [1913]

    doi:10.24963/ijcai.2019/264

    2019 doi

  44. [2020]

    IEEE Transactions on Intelligent Transportation Systems 23, 2 (2020), 1009–1018

    Dynamic graph convolution network for traffic forecasting based on latent network of laplace matrix estimation. IEEE Transactions on Intelligent Transportation Systems 23, 2 (2020), 1009–1018

    2020

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.