Pith. sign in

REVIEW 5 major objections 5 minor 29 references

Federated Learning with Graph-Based Aggregation for Traffic Forecasting

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read Two lightweight graph-aware aggregation rules for federated traffic forecasting beat FedAvg, FMTL, and CNFGNN on METR-LA and PEMS-BAY, the paper reports.

desk verdict A clean lightweight aggregation idea, but the comparison to heavier baselines is not apples-to-apples; the claimed 1.9-8.1% gains are not established by the current experiments. read the letter →

arxiv 2507.09805 v1 pith:UZ5UBCLJ submitted 2025-07-13 cs.LG cs.AI

classification cs.LGcs.AI
keywords federatedlearningtrafficforecastinggraph-basedaggregationaveraginglabelpropagationspatio-temporalpredictionMETR-LAPEMS-BAY
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 server-side graph-aware aggregation for federated traffic forecasting does not need a graph neural network: two simple weighted-averaging rules that use the road-network graph can outperform standard federated averaging and more complex graph-based methods. On the METR-LA and PEMS-BAY benchmarks, the authors report that GraphFedAvg and MPFedAvg reach lower root mean squared error than FedAvg, FMTL, and CNFGNN, with improvements of 1.9% to 8.1% over the next-best method, while adding only a matrix multiplication to the server's aggregation step. If the claim holds, resource-constrained federated systems could capture spatial dependencies between sensors almost for free, without training deep graph models on the server.

What carries the argument

The mechanism is a server-side replacement of uniform averaging by graph propagation over the road-network adjacency matrix $A$, augmented with self-loops as $\tilde{A}=A+I$. GraphFedAvg iterates $X^{(\ell+1)}=\tilde{D}^{-1}\tilde{A}X^{(\ell)}$ for $L$ steps, where $\tilde{D}$ is the degree matrix of $\tilde{A}$; this makes each client's updated parameters a degree-normalized average of parameters from its $L$-hop neighbourhood. MPFedAvg instead iterates the label-propagation-style rule $X^{(\ell+1)}=\alpha\,\tilde{D}^{-1/2}\tilde{A}\tilde{D}^{-1/2}X^{(\ell)}+(1-\alpha)X^{(\ell)}$, with $\alpha=0.8$, blending a symmetrically normalized neighbor average with each client's own parameters. Both rules are linear in the parameter matrix, so they keep the aggregation step close to FedAvg in cost while letting information flow across the sensor graph.

What would settle it

A degree-preserving random permutation of the adjacency matrix, rerun through both aggregation rules, would settle the central mechanism: if RMSE does not rise substantially, the road graph's spatial structure is not what produces the reported gains.

Watch

Extended reading notes

Core claim

The central claim, stated by the authors in the results section, is that replacing the uniform average of FedAvg with a neighbourhood average over the road graph yields the best forecasting accuracy among all compared methods on both datasets. The best single-layer results are RMSE 3.733 on PEMS-BAY for MPFedAvg and 11.473 on METR-LA for GraphFedAvg, against 3.822 and 11.487 for CNFGNN, the strongest baseline. The authors conclude that a single propagation step captures the essential spatial dependencies, that a second step adds almost nothing, and that the gain comes from preserving inter-client dependencies without server-side GNN training.

Load-bearing premise

The load-bearing premise is that the sensor-distance road graph is a faithful map of which clients' model parameters should be averaged together; if the graph does not track parameter similarity, the neighbourhood weighting can only distort the global model.

Editorial extensions

If this is right

  • If the reported results hold, a single propagation step is enough for most of the benefit, so the one-layer variants offer the best accuracy-to-cost trade-off.
  • The server can skip GNN training entirely; the entire graph-aware component is a few sparse matrix multiplications over client parameters.
  • The privacy properties of standard federated learning are preserved, because the server still receives only model updates, never raw traffic readings.
  • The same aggregation recipe is directly reusable in any federated spatio-temporal task where a meaningful client-similarity graph exists.

Reading between the lines

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

  • An implication the authors leave implicit is that the graph's role may be regularization rather than new information: a degree-preserving random rewire of the adjacency matrix should erase most of the reported gain if the road structure is doing the work.
  • A testable extension beyond the paper is to replace the static road graph with a graph learned from parameter similarity each round; this would show whether the gains require physical proximity or merely a better averaging weight.
  • Because the aggregation rules are agnostic to the local model, the same gains may transfer to clients using architectures other than GRUs; the current experiments do not vary the client model.
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

5 major / 5 minor

Summary. The paper proposes two lightweight graph-aware federated averaging methods, GraphFedAvg and MPFedAvg, for traffic forecasting. In GraphFedAvg, the server replaces FedAvg's uniform averaging with a normalized averaging over each client's graph neighbors; in MPFedAvg, a label-propagation-style update blends each client's parameters with a degree-normalized neighborhood average controlled by a scalar α. The authors argue that these methods capture spatial dependencies between clients without the computational cost of full server-side GNN training. The experimental section compares the methods against GRU-based centralized, local, and FedAvg baselines, as well as FMTL and CNFGNN, on METR-LA and PEMS-BAY, reporting RMSE values. The central claim in Section 6 is that the proposed methods outperform all baselines on both datasets with improvements of 1.9% to 8.1% over CNFGNN.

Significance. If the experimental claims were fully supported, the paper would offer a simple and computationally attractive alternative to graph-neural-network-based federated aggregation for spatio-temporal forecasting. The aggregation rules in Sections 4.1 and 4.2 are clearly specified and easy to implement, and they are a reasonable extension of FedAvg that exploits a known client graph. The paper also correctly identifies that existing server-side graph approaches introduce substantial complexity. However, the current experimental validation is not sufficient to establish the central claim: the comparison against the two strongest baselines relies on numbers taken from a different experimental protocol, no repeated runs are reported, the metric set is incomplete, and one variant actually performs worse than CNFGNN on METR-LA in Table 1. The contribution is potentially useful, but the evidence presented here is not yet convincing.

major comments (5)
  1. [Section 6, Table 1] The central claim that 'our proposed methods outperform all baselines across both datasets' is directly contradicted by Table 1 for MPFedAvg (1L) on METR-LA: the reported RMSE is 11.489, which is higher than the CNFGNN value of 11.487. Even if this small difference were within noise, the sentence as written is false for at least one proposed variant. The claim should be restricted to the variants that actually improve over CNFGNN, and the comparison needs statistical support.
  2. [Section 5.3 and Table 1 footnote] The comparison against FMTL and CNFGNN is not apples-to-apples. The footnote to Table 1 states that FMTL and CNFGNN results are '* Results as provided by [15]', and only GRU (centralized), GRU (local), and GRU + FedAvg were re-run. Section 5.3 fixes the protocol for the proposed methods to 3 local epochs and 5 client-server rounds with Adam lr=1e-3. The published CNFGNN and FMTL numbers in [15] come from a different training budget and protocol. Therefore the claimed 1.9%–8.1% improvement over CNFGNN may be a comparison artifact rather than a property of the proposed aggregation rules. The authors must either re-run these baselines under the identical protocol or substantially qualify the claim.
  3. [Table 1 and Section 5.1] The experimental evidence lacks statistical grounding. Only single RMSE values are reported, with no error bars, standard deviations, or repeated seeds, even though Section 5.1 says three metrics (MAE, MAPE, RMSE) are used and only RMSE is shown. The METR-LA differences between the best proposed variant (11.473) and CNFGNN (11.487) are about 0.1%, which is well within typical seed-to-seed variation for GRU training; without repeated runs this cannot support an 'outperform' claim. The authors should report mean and variance over multiple seeds for all methods, including the re-run baselines.
  4. [Section 6, first two paragraphs] There is an internal inconsistency in the baseline narrative. The paper states that centralized GRU (4.172 on PEMS-BAY) 'serves as an upper bound' and that local training shows weaker results, but Table 1 reports GRU (local) at 4.152, which is better than centralized GRU on PEMS-BAY. This contradiction suggests either a protocol difference, a typo, or an issue with how the baselines were run, and it must be explained before the baseline discussion can be trusted.
  5. [Sections 4.1, 4.2 and 5.3] The two hyperparameters of the proposed methods, α=0.8 for MPFedAvg and the propagation depth L for both methods, are fixed by hand with no sensitivity analysis or validation-curve evidence. Since the reported gains are small on METR-LA, the possibility that these choices were tuned on the test set is a real concern. The authors should provide a sensitivity study over α and L, and an ablation with an alternative graph (e.g., random or learned) to justify that the road-network adjacency, not merely the averaging operation, is responsible for the improvements.
minor comments (5)
  1. [Section 2, 'and and'] There is a typo in the sentence 'each client maintains its own local data (spatial and temporal) and and trains its local models'; the duplicated 'and' should be removed.
  2. [Section 5.3] The name 'LPFedAvg' appears in the sentence 'For the Graph Message Passing Aware (LPFedAvg) model we set α = 0.8', but the method is called MPFedAvg in Sections 4.2, 6, and the abstract. The naming should be made consistent.
  3. [Section 3] The adjacency matrix A is defined as binary (A_ij ∈ {0,1}), while Section 5.1 says the adjacency is constructed with a thresholded Gaussian kernel over road-network distances, which typically produces real-valued weights. Please clarify whether the aggregation uses the binary adjacency or the weighted kernel values.
  4. [Section 5.1 and Table 1] The paper says experiments evaluate MAE, MAPE, and RMSE, but Table 1 reports only RMSE. Either the other metrics should be reported or the text should be adjusted to state that RMSE is the primary reported metric.
  5. [Section 7] The conclusion repeats the '1.9% to 8.1%' improvement claim without acknowledging the caveat that FMTL and CNFGNN numbers are taken from a different protocol. Please qualify this claim in the conclusion as well.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the aggregation rules are standalone operations and the reported RMSE values are empirical outputs, not algebraic consequences of the inputs.

full rationale

The paper's central derivation is the definition of two server-side aggregation rules: GraphFedAvg as X^(l+1) = D~^-1 A~ X^(l) and MPFedAvg as X^(l+1) = alpha D~^-1/2 A~ D~^-1/2 X^(l) + (1-alpha) X^(l). These are explicitly stated update rules over client parameter matrices; they are not derived from, nor equivalent to, the reported RMSE values in Table 1. No fitted parameter is renamed as a prediction: the only hyperparameter, alpha = 0.8, is set by hand and the propagation depth L is a chosen model variant, not fitted to the test RMSE. The graph is taken from the previously published thresholded Gaussian kernel construction cited to [15], and the paper does not claim to derive that graph from its own results. The comparison to FMTL and CNFGNN relies on RMSE values 'as provided by [15]' rather than re-runs under the paper's 5-round protocol; that is a genuine concern about comparison fairness and statistical strength, but it is not circularity because the claimed improvements are empirical measurements against external published numbers, not consequences of the paper's own definitions. There is no self-citation chain: the authors cite prior work by other groups for baselines, graph construction, and algorithms. Therefore the derivation is self-contained with respect to the circularity patterns considered; score 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim rests on two domain assumptions: that the road-network graph is a good guide for parameter similarity, and that averaging GRU parameters across non-IID clients is meaningful. The only tuned model hyperparameter is alpha (0.8) and the propagation depth (1 or 2). No new entities are introduced.

free parameters (2)
  • alpha (α) in MPFedAvg = 0.8
    Chosen hyperparameter controlling the blend of self-parameters and neighbor parameters; set without reported sensitivity analysis (Section 5.3).
  • Propagation depth L = 1 and 2
    The number of neighbor aggregation hops is evaluated at 1 and 2 layers, and both are reported; depth is selected per variant and affects the results (Section 4, Table 1).
assumptions (3)
  • domain assumption The client graph adjacency matrix from road-network distances is a meaningful guide for parameter aggregation.
    Both methods weight client updates by A (Sections 4.1 and 4.2), with A built via a thresholded Gaussian kernel in Section 5.1; the paper assumes spatial closeness implies parameter-update similarity.
  • domain assumption Local client models can be averaged in parameter space despite non-IID data.
    The methods aggregate GRU encoder-decoder parameters directly (Section 4), inheriting FedAvg's assumption that parameter-space averaging produces a useful global model when local objectives differ (Section 5.3).
  • standard math Matrix multiplications and symmetric normalization behave as in standard graph convolution.
    The update rules in Sections 4.1 and 4.2 are standard normalized adjacency operations; no proof of convergence is given, so convergence is assumed from FedAvg-style training.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Federated Learning with Graph-Based Aggregation for Traffic Forecasting." pith.science (2026). https://pith.science/paper/UZ5UBCLJ

@misc{pith2026250709805,
  author       = {Pith},
  title        = {Pith review of: Federated Learning with Graph-Based Aggregation for Traffic Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UZ5UBCLJ}},
  note         = {Machine review of arXiv:2507.09805}
}
read the original abstract

In traffic prediction, the goal is to estimate traffic speed or flow in specific regions or road segments using historical data collected by devices deployed in each area. Each region or road segment can be viewed as an individual client that measures local traffic flow, making Federated Learning (FL) a suitable approach for collaboratively training models without sharing raw data. In centralized FL, a central server collects and aggregates model updates from multiple clients to build a shared model while preserving each client's data privacy. Standard FL methods, such as Federated Averaging (FedAvg), assume that clients are independent, which can limit performance in traffic prediction tasks where spatial relationships between clients are important. Federated Graph Learning methods can capture these dependencies during server-side aggregation, but they often introduce significant computational overhead. In this paper, we propose a lightweight graph-aware FL approach that blends the simplicity of FedAvg with key ideas from graph learning. Rather than training full models, our method applies basic neighbourhood aggregation principles to guide parameter updates, weighting client models based on graph connectivity. This approach captures spatial relationships effectively while remaining computationally efficient. We evaluate our method on two benchmark traffic datasets, METR-LA and PEMS-BAY, and show that it achieves competitive performance compared to standard baselines and recent graph-based federated learning techniques.

Figures

Figures reproduced from arXiv: 2507.09805 by the authors.

Figure 1
Figure 1. Illustration of a traffic forecasting system in a centralized FL system. Each road intersection is equipped with a [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 14 canonical work pages

  1. [15]

    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

  2. [1]

    Rasha Al-Huthaifi, Tianrui Li, Zaid Al-Huda, and Chongshou Li. 2024. FedAGAT: Real-time traffic flow prediction based on federated community and adaptive graph attention network. Information Sciences 667 (2024), 120482. https://doi. org/10.1016/j.ins.2024.120482

  3. [2]

    Kyunghyun Cho, Bart Van Merriënboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078 (2014)

  4. [3]

    Liam Collins, Hamed Hassani, Aryan Mokhtari, and Sanjay Shakkottai. 2021. Exploiting shared representations for personalized federated learning. In Inter- national conference on machine learning . PMLR, 2089–2099

  5. [4]

    Hejie Cui, Wei Dai, Yanqiao Zhu, Xuan Kan, Antonio Aodong Chen Gu, Joshua Lukemire, Liang Zhan, Lifang He, Ying Guo, and Carl Yang. 2023. BrainGB: A Benchmark for Brain Network Analysis With Graph Neural Networks. IEEE Transactions on Medical Imaging 42, 2 (2023), 493–506. https://doi.org/10.1109/ TMI.2022.3218745

  6. [5]

    Jian Feng, Cailing Du, and Qi Mu. 2024. Traffic Flow Prediction Based on Federated Learning and Spatio-Temporal Graph Neural Networks. ISPRS International Journal of Geo-Information 13 (06 2024), 210. https://doi.org/10.3390/ijgi13060210

  7. [6]

    Xinxin Feng, Haoran Sun, Shunjian Liu, Junxin Guo, and Haifeng Zheng. 2024. Federated Meta-Learning on Graph for Traffic Flow Prediction. IEEE Transactions on Vehicular Technology 73, 12 (2024), 19526–19538. https://doi.org/10.1109/TVT. 2024.3441759

  8. [7]

    Yang, Yongdong Shen, Panagiotis Angeloudis, Leandro Parada, and Chao Wu

    Simon Hu, Yin Ye, Qinru Hu, Xin Liu, Shaosheng Cao, Howard H. Yang, Yongdong Shen, Panagiotis Angeloudis, Leandro Parada, and Chao Wu. 2023. A Federated Learning-Based Framework for Ride-Sourcing Traffic Demand Prediction. IEEE Transactions on Vehicular Technology 72, 11 (2023), 14002–14015. https://doi.org/ 10.1109/TVT.2023.3287221

Show all 29 references
  1. [8]

    Kaiyuan Li, Yihan Zhang, Huandong Wang, Yan Zhuo, and Xinlei Chen. 2024. FedASTA: Federated adaptive spatial-temporal attention for traffic flow predic- tion. arXiv preprint arXiv:2405.13090 (2024)

  2. [9]

    Tian Li, Anit Kumar Sahu, Manzil Zaheer, Maziar Sanjabi, Ameet Talwalkar, and Virginia Smith. 2020. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems 2 (2020), 429–450

  3. [10]

    Yi Li, Renyou Xie, Chaojie Li, Yi Wang, and Zhaoyang Dong. 2024. Federated Graph Learning for EV Charging Demand Forecasting with Personalization Against Cyberattacks. arXiv preprint arXiv:2405.00742 (2024)

  4. [11]

    Yaguang Li, Rose Yu, Cyrus Shahabi, and Yan Liu. 2017. Diffusion convolu- tional recurrent neural network: Data-driven traffic forecasting. arXiv preprint arXiv:1707.01926 (2017)

  5. [12]

    Lei Liu, Yuxing Tian, Chinmay Chakraborty, Jie Feng, Qingqi Pei, Li Zhen, and Keping Yu. 2023. Multilevel Federated Learning-Based Intelligent Traffic Flow Forecasting for Transportation Network Management. IEEE Transactions on Network and Service Management 20, 2 (2023), 1446...

  6. [13]

    Ruyue Liu, Rong Yin, Xiangzhen Bo, Xiaoshuai Hao, Xingrui Zhou, Yong Liu, Can Ma, and Weiping Wang. 2024. Communication-Efficient Personalized Federal Graph Learning via Low-Rank Decomposition. arXiv preprint arXiv:2412.13442 (2024)

  7. [14]

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep net- works from decentralized data. In Artificial intelligence and statistics . PMLR, 1273–1282

  8. [16]

    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. Appl. Soft Comput. 138, C (May 2023), 11 pages. https://doi.org/10.1016/j.asoc.2023.110175

  9. [17]

    Arash Rasti-Meymandi, Seyed Mohammad Sheikholeslami, Jamshid Abouei, and Konstantinos N Plataniotis. 2022. Graph federated learning for CIoT devices in smart home applications. IEEE Internet of Things Journal 10, 8 (2022), 7062–7079

  10. [18]

    Virginia Smith, Chao-Kai Chiang, Maziar Sanjabi, and Ameet S Talwalkar. 2017. Federated multi-task learning. Advances in neural information processing systems 30 (2017)

  11. [19]

    Yuxing Tian, Jiachi Luo, Zheng Liu, Song Li, and Yanwen Qu. 2023. M 3 FGM: A Node Masking and Multi-granularity Message Passing-Based Federated Graph Model for Spatial-Temporal Data Prediction. InInternational Conference on Neural Information Processing. Springer, 551–566

  12. [20]

    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). 221–225. https://doi.org/10....

  13. [21]

    Han Xie, Yi Yang, Hejie Cui, and Carl Yang. 2024. Federated learning for cross- institution brain network analysis. In Medical Imaging 2024: Computer-Aided Diagnosis, Vol. 12927. SPIE, 106–119

  14. [22]

    Xiaoming Yuan, Jiahui Chen, Jiayu Yang, Ning Zhang, Tingting Yang, Tao Han, and Amir Taherkordi. 2023. FedSTN: Graph Representation Driven Federated Learning for Edge Computing Enabled Urban Traffic Flow Prediction. IEEE Transactions on Intelligent Transportation Systems 24, 8...

  15. [23]

    Chuanting Zhang, Shuping Dang, Basem Shihada, and Mohamed-Slim Alouini

  16. [24]

    Yu Zhang, Hua Lu, Ning Liu, Yonghui Xu, Qingzhong Li, and Lizhen Cui. 2024. Personalized federated learning for cross-city traffic prediction. In 33rd Interna- tional Joint Conference on Artificial Intelligence, IJCAI . 5526–5534

  17. [25]

    Yu Zhang, Hua Lu, Ning Liu, Yonghui Xu, Qingzhong Li, and Lizhen Cui. 2024. Personalized Federated Learning for Cross-City Traffic Prediction. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJCAI- 24, Kate Larson (Ed.). Internatio...

  18. [26]

    Yudong Zhang, Xu Wang, Xuan Yu, Kuo Yang, Zhengyang Zhou, and Yang Wang

  19. [27]

    Xiaojin Zhur and Zoubin Ghahramanirh. 2003. Learning from labeled and unla- beled data with label propagation. (2003)

  20. [2021]

    In IEEE INFOCOM 2021-IEEE conference on computer communications

    Dual attention-based federated learning for wireless traffic prediction. In IEEE INFOCOM 2021-IEEE conference on computer communications . IEEE, 1–10

  21. [2025]

    In Companion Proceedings of the ACM on Web Conference 2025 (Sydney NSW, Australia) (WWW ’25)

    FedSTG: Breaking through Spatio-Temporal Data Silos with Federated Graph Learning. In Companion Proceedings of the ACM on Web Conference 2025 (Sydney NSW, Australia) (WWW ’25) . Association for Computing Machinery, New York, NY, USA, 1534–1538. https://doi.org/10.1145/3701716.3715562

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.