pith. sign in

arxiv: 2503.15845 · v2 · submitted 2025-03-20 · 💻 cs.LG

Network-wide Freeway Traffic Estimation Using Sparse Sensor Data: A Dirichlet Graph Auto-Encoder Approach

Qishen Zhou , Yifan Zhang , Michail A. Makridis , Anastasios Kouvelas , Yibing Wang , Simon Hu This is my paper

Pith reviewed 2026-05-22 23:34 UTC · model grok-4.3

classification 💻 cs.LG
keywords traffic state estimationgraph auto-encodersparse sensorsdirected graphsDirichlet energy propagationphysics-guided modelingfreeway trafficcross-city transfer
0
0 comments X

The pith

A directed-graph auto-encoder estimates complete freeway traffic states from sparse sensors by propagating features without zero-filling and by modeling congestion and free-flow separately.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Network-wide traffic state estimation must fill in conditions at sensor-free locations, yet standard graph methods introduce bias by replacing missing values with zeros before message passing. The paper develops DGAE, an inductive model that first derives a directed version of Dirichlet energy feature propagation (DEFP4D) to avoid that bias, then encodes latent representations guided by this propagation and applies separate physics-guided rules for congested versus free-flow regimes. Experiments on three freeway datasets show the model exceeds prior state-of-the-art accuracy while transferring across cities without retraining. DEFP4D alone also performs well when sensors are extremely sparse. These results matter because they reduce the sensor density required for reliable network-level traffic pictures used in management and control.

Core claim

The paper establishes that its DGAE model, which combines theoretically derived DEFP4D for directed graphs, DEFP4D-guided latent space encoding, and physics-guided propagation mechanisms that treat congested and free-flow patterns separately, produces more accurate network-wide traffic estimates than existing methods under sparse sensor conditions and demonstrates cross-city transferability.

What carries the argument

DEFP4D-guided latent space encoding inside a graph auto-encoder that applies separate physics-guided propagation for congested and free-flow patterns on directed traffic networks.

If this is right

  • DGAE outperforms existing state-of-the-art methods on three traffic datasets.
  • The model exhibits strong cross-city transferability without retraining.
  • DEFP4D alone functions as a lightweight solution that remains accurate under extremely sparse sensor conditions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same directed propagation mechanism could be tested on non-freeway networks where flow directions are less strictly one-way.
  • Because the method avoids labeled data for the separate-pattern modeling step, it may reduce supervision needs in other sparse-graph imputation tasks.
  • If DEFP4D remains effective when the underlying graph changes topology, the approach could support real-time sensor reconfiguration without full model retraining.

Load-bearing premise

Distinct propagation rules for congested and free-flow patterns can be modeled separately via physics-guided mechanisms without requiring additional labeled data or introducing bias in the directed graph setting.

What would settle it

On a held-out freeway dataset or under a new sparsity level, DGAE fails to exceed the accuracy of the strongest prior method or loses cross-city transfer performance.

Figures

Figures reproduced from arXiv: 2503.15845 by Anastasios Kouvelas, Michail A. Makridis, Qishen Zhou, Simon Hu, Yibing Wang, Yifan Zhang.

Figure 1
Figure 1. Figure 1: Illustration of foundational framework Dirichlet graph auto-encoder. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of bidirectional diffusion dynamics of traffic flow: an [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of the refined DGAE architecture. The Graph Encoder structure, identical to the Graph Decoder, is omitted here for brevity. MLP denotes [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: State estimation results for METR-LA: (a) Sensor #1, (b) Sensor #70, (c) Sensor #131. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Model performance with different sensor deployment density on METR-LA dataset. The first row of x-ticks labels indicates the number of VS, while [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The effectiveness of DEFP4D in latent space representation. We compare our proposed DGAE with a variant that applied DEFP4D in the original [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: A map view of estimation error distribution. The uncolored and colored circles indicate AS and VS, respectively, where different colors indicate [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
read the original abstract

Network-wide Traffic State Estimation (TSE), which aims to infer a complete image of network traffic states with sparsely deployed sensors, plays a vital role in intelligent transportation systems. With the development of data-driven methods, traffic dynamics modeling has advanced significantly. However, TSE poses fundamental challenges for data-driven approaches, since historical patterns cannot be learned locally at sensor-free segments. Although graph representation learning shows promise in estimating states at locations without sensors, existing methods typically handle unobserved locations by filling them with zeros, introducing bias to the sensitive graph message propagation. The recently proposed Dirichlet Energy-based Feature Propagation (DEFP) method achieves State-Of-The-Art (SOTA) performance in unobserved node classification by eliminating the need for zero-filling. However, applying it to TSE faces three key challenges: inability to handle directed traffic networks, strong assumptions in traffic spatial correlation modeling, and overlooking distinct propagation rules of different patterns (e.g., congestion and free flow). We propose DGAE, a novel inductive graph representation model that addresses these challenges through theoretically derived DEFP for Directed graph (DEFP4D), enhanced spatial representation learning via DEFP4D-guided latent space encoding, and physics-guided propagation mechanisms that separately handle congested and free-flow patterns. Experiments on three traffic datasets demonstrate that DGAE outperforms existing SOTA methods and exhibits strong cross-city transferability. Furthermore, DEFP4D can serve as a standalone lightweight solution, showing superior performance under extremely sparse sensor conditions. The code of this work is publicly available at: https://github.com/ZJU-TSELab/DGAE.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes DGAE, a Dirichlet Graph Auto-Encoder for network-wide freeway traffic state estimation (TSE) from sparse sensors on directed graphs. It introduces DEFP4D (theoretically derived Dirichlet Energy-based Feature Propagation for Directed graphs) to avoid zero-filling bias, uses DEFP4D-guided latent space encoding for spatial representation, and adds physics-guided propagation mechanisms that separately model congested and free-flow patterns. Experiments on three traffic datasets claim outperformance over SOTA methods, strong cross-city transferability, and that DEFP4D works as a lightweight standalone solution under extreme sparsity. Code is released publicly.

Significance. If the central claims hold, the work would advance data-driven TSE by providing a principled way to handle directed networks and pattern-specific dynamics without zero-filling or extra labels, with practical value for sparse-sensor freeway monitoring and transfer across cities. The public code release strengthens reproducibility.

major comments (3)
  1. [§4.3] §4.3 (Physics-guided propagation mechanisms): The claim that congested and free-flow patterns can be handled separately via physics-guided rules without additional labeled data or bias is load-bearing for the headline performance gains under sparsity. The manuscript must explicitly show the regime-assignment procedure from sparse directed observations alone and demonstrate (via ablation or sensitivity analysis) that it does not correlate with edge directions or introduce bias precisely where sensors are absent.
  2. [§5] §5 (Experiments, cross-city transfer results): The reported outperformance and transferability depend on the DEFP4D + physics-guided components. The tables must include ablation removing the pattern-separation step and report the exact sensor sparsity levels, directed-graph construction details, and whether any auxiliary regime labels were used during training or inference.
  3. [§3.2] §3.2 (DEFP4D derivation): The extension from undirected DEFP to directed graphs is central; the manuscript should verify that the energy functional and message-passing rules remain parameter-free and do not implicitly reintroduce zero-filling or direction-dependent bias when applied to traffic networks.
minor comments (2)
  1. [Abstract] Abstract and §1: The three addressed challenges are clearly stated, but the transition from DEFP limitations to the proposed solutions would benefit from one sentence on how DEFP4D avoids the zero-filling issue in directed settings.
  2. Notation: Ensure consistent use of symbols for directed adjacency and the energy functional across §3 and §4.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback and detailed comments. We address each major comment point-by-point below. Where revisions are needed for clarity or additional analysis, we will incorporate them in the next version of the manuscript.

read point-by-point responses

  1. Referee: [§4.3] §4.3 (Physics-guided propagation mechanisms): The claim that congested and free-flow patterns can be handled separately via physics-guided rules without additional labeled data or bias is load-bearing for the headline performance gains under sparsity. The manuscript must explicitly show the regime-assignment procedure from sparse directed observations alone and demonstrate (via ablation or sensitivity analysis) that it does not correlate with edge directions or introduce bias precisely where sensors are absent.

    Authors: We agree that explicit documentation of the regime-assignment procedure and supporting analyses are required. The procedure classifies regimes from sparse observations alone using a speed-threshold rule grounded in fundamental traffic flow diagrams (free-flow above a calibrated threshold, congested below), applied only to observed nodes before propagation. In the revision we will add a dedicated paragraph in §4.3 describing this step-by-step, together with an ablation that removes pattern separation and a sensitivity study checking correlation with edge direction and missing-sensor locations. These additions will directly address the load-bearing claim. revision: yes

  2. Referee: [§5] §5 (Experiments, cross-city transfer results): The reported outperformance and transferability depend on the DEFP4D + physics-guided components. The tables must include ablation removing the pattern-separation step and report the exact sensor sparsity levels, directed-graph construction details, and whether any auxiliary regime labels were used during training or inference.

    Authors: We accept the request for fuller reporting. The revised §5 will add a new ablation column/table that isolates the effect of removing the pattern-separation step, list the precise sparsity ratios tested (5 %, 10 %, 20 %, 30 %), describe the directed-graph construction (nodes as freeway segments, directed edges following observed traffic flow direction from the network topology), and explicitly state that no auxiliary regime labels were supplied at any stage. These details will also be noted in the cross-city transfer experiments. revision: yes

  3. Referee: [§3.2] §3.2 (DEFP4D derivation): The extension from undirected DEFP to directed graphs is central; the manuscript should verify that the energy functional and message-passing rules remain parameter-free and do not implicitly reintroduce zero-filling or direction-dependent bias when applied to traffic networks.

    Authors: We will strengthen §3.2 with an explicit verification subsection. The directed energy functional is obtained by replacing the symmetric Laplacian with the directed combinatorial Laplacian while preserving the quadratic form; the resulting message-passing rule stays parameter-free and operates only on observed features, thereby avoiding zero-filling. We will include the algebraic steps showing that no direction-dependent bias is introduced beyond the topology itself, plus a short empirical check on the traffic graphs confirming the absence of reintroduced zero-filling artifacts. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper derives DEFP4D for directed graphs and introduces physics-guided mechanisms for congested vs. free-flow patterns, then validates DGAE empirically on three traffic datasets with reported outperformance and cross-city transfer. No equations, fitted parameters, or self-citations are shown in the provided text that reduce any claimed prediction, uniqueness result, or performance gain to an input by construction. The experimental claims and inductive model remain independent of the derivation steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review performed on abstract only; no explicit free parameters, axioms, or invented entities are stated in the provided text.

pith-pipeline@v0.9.0 · 5844 in / 1077 out tokens · 37271 ms · 2026-05-22T23:34:09.663792+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

59 extracted references · 59 canonical work pages

  1. [1]

    State estimation of a physical system with unknown governing equations,

    K. Course and P. B. Nair, “State estimation of a physical system with unknown governing equations,”Nature, vol. 622, no. 7982, pp. 261–267, Oct. 2023

    work page 2023

  2. [2]

    Traffic state estimation on highway: A comprehensive survey,

    T. Seo, A. M. Bayen, T. Kusakabe, and Y . Asakura, “Traffic state estimation on highway: A comprehensive survey,” Annual Reviews in Control, vol. 43, pp. 128–151, 2017

    work page 2017

  3. [3]

    An extended Kalman filter approach for real-time state estimation in multi-region MFD urban networks,

    M. Saeedmanesh, A. Kouvelas, and N. Geroliminis, “An extended Kalman filter approach for real-time state estimation in multi-region MFD urban networks,” Transportation Research Part C: Emerging Technologies, vol. 132, p. 103384, Nov. 2021

    work page 2021

  4. [4]

    Real-time joint traffic state and model parameter estimation on freeways with fixed sensors and connected vehicles: State-of-the-art overview, methods, and case studies,

    Y . Wang, M. Zhao, X. Yu, Y . Hu, P. Zheng, W. Hua, L. Zhang, S. Hu, and J. Guo, “Real-time joint traffic state and model parameter estimation on freeways with fixed sensors and connected vehicles: State-of-the-art overview, methods, and case studies,” Transportation Research Part C: Emerging Technologies, vol. 134, p. 103444, 2022

    work page 2022

  5. [5]

    An adaptive framework for real- time freeway traffic estimation in the presence of cavs,

    M. A. Makridis and A. Kouvelas, “An adaptive framework for real- time freeway traffic estimation in the presence of cavs,” Transportation Research Part C: Emerging Technologies, vol. 149, p. 104066, 2023

    work page 2023

  6. [6]

    Stochastic Switching Mode Model based Filters for urban arterial traffic estimation from multi-source data,

    X.-S. Trinh, M. Keyvan-Ekbatani, D. Ngoduy, and B. Robertson, “Stochastic Switching Mode Model based Filters for urban arterial traffic estimation from multi-source data,” Transportation Research Part C: Emerging Technologies, vol. 164, p. 104664, July 2024

    work page 2024

  7. [7]

    Bayesian kernelized matrix factorization for spatiotemporal traffic data imputation and kriging,

    M. Lei, A. Labbe, Y . Wu, and L. Sun, “Bayesian kernelized matrix factorization for spatiotemporal traffic data imputation and kriging,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 10, pp. 18 962–18 974, 2022

    work page 2022

  8. [8]

    Correlating sparse sensing for large-scale traffic speed estimation: A Laplacian-enhanced low-rank tensor kriging approach,

    T. Nie, G. Qin, Y . Wang, and J. Sun, “Correlating sparse sensing for large-scale traffic speed estimation: A Laplacian-enhanced low-rank tensor kriging approach,” Transportation Research Part C: Emerging Technologies, vol. 152, p. 104190, July 2023

    work page 2023

  9. [9]

    GE-GAN: A novel deep learning framework for road traffic state estimation,

    D. Xu, C. Wei, P. Peng, Q. Xuan, and H. Guo, “GE-GAN: A novel deep learning framework for road traffic state estimation,” Transportation Research Part C-Emerging Technologies , vol. 117, p. 102635, Aug. 2020

    work page 2020

  10. [10]

    C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning. Cambridge, MA, USA: The MIT Press, Nov. 2005

    work page 2005

  11. [11]

    Cressie and C

    N. Cressie and C. K. Wikle, Statistics for spatio-temporal data . John Wiley & Sons, 2011

    work page 2011

  12. [12]

    Neural message passing for quantum chemistry,

    J. Gilmer, S. S. Schoenholz, P. F. Riley, O. Vinyals, and G. E. Dahl, “Neural message passing for quantum chemistry,” in International conference on machine learning . PMLR, 2017, pp. 1263–1272

    work page 2017

  13. [13]

    Inductive graph neural networks for spatiotemporal kriging,

    Y . Wu, D. Zhuang, A. Labbe, and L. Sun, “Inductive graph neural networks for spatiotemporal kriging,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 35, no. 5, 2021, pp. 4478– 4485

    work page 2021

  14. [14]

    AGNP: Network- wide short-term probabilistic traffic speed prediction and imputation,

    M. Xu, Y . Di, H. Ding, Z. Zhu, X. Chen, and H. Yang, “AGNP: Network- wide short-term probabilistic traffic speed prediction and imputation,” Communications in Transportation Research , vol. 3, p. 100099, Dec. 2023

    work page 2023

  15. [15]

    Kits: Inductive spatio-temporal kriging with increment training strategy,

    Q. Xu, C. Long, Z. Li, S. Ruan, R. Zhao, and Z. Li, “Kits: Inductive spatio-temporal kriging with increment training strategy,”

    work page

  16. [16]

    Available: https://arxiv.org/abs/2311.02565

    [Online]. Available: https://arxiv.org/abs/2311.02565

    work page arXiv

  17. [17]

    Towards better traffic volume es- timation: Jointly addressing the underdetermination and nonequilibrium problems with correlation-adaptive GNNs,

    T. Nie, G. Qin, Y . Wang, and J. Sun, “Towards better traffic volume es- timation: Jointly addressing the underdetermination and nonequilibrium problems with correlation-adaptive GNNs,” Transportation Research Part C: Emerging Technologies, vol. 157, p. 104402, Dec. 2023

    work page 2023

  18. [18]

    Multi-view graph imputation network,

    X. Peng, J. Cheng, X. Tang, B. Zhang, and W. Tu, “Multi-view graph imputation network,” Information Fusion , vol. 102, p. 102024, Feb. 2024. 13

    work page 2024

  19. [19]

    Why Not to Use Zero Imputation? Correcting Sparsity Bias in Training Neural Networks,

    J. Yi, J. Lee, K. J. Kim, S. J. Hwang, and E. Yang, “Why Not to Use Zero Imputation? Correcting Sparsity Bias in Training Neural Networks,” in International Conference on Learning Representations , Sept. 2019

    work page 2019

  20. [20]

    On the unreasonable effectiveness of feature propagation in learning on graphs with missing node features,

    E. Rossi, H. Kenlay, M. I. Gorinova, B. P. Chamberlain, X. Dong, and M. M. Bronstein, “On the unreasonable effectiveness of feature propagation in learning on graphs with missing node features,” in Learning on graphs conference . PMLR, 2022, pp. 11–1

    work page 2022

  21. [21]

    On kinematic waves ii. a theory of traffic flow on long crowded roads,

    M. J. Lighthill and G. B. Whitham, “On kinematic waves ii. a theory of traffic flow on long crowded roads,” Proc. Roy. Soc. London A. Math. Phys. Sci., vol. 229, no. 1178, pp. 317–345, 1955

    work page 1955

  22. [22]

    High time-resolution queue profile estimation at signalized intersections based on extended kalman filtering,

    S. Hu, Q. Zhou, J. Li, Y . Wang, C. Roncoli, L. Zhang, and L. Lehe, “High time-resolution queue profile estimation at signalized intersections based on extended kalman filtering,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 11, pp. 21 274–21 290, 2022

    work page 2022

  23. [23]

    Physics-informed deep learning for traffic state estimation: A hybrid paradigm informed by second-order traffic models,

    R. Shi, Z. Mo, and X. Di, “Physics-informed deep learning for traffic state estimation: A hybrid paradigm informed by second-order traffic models,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 35, no. 1, pp. 540–547, May 2021

    work page 2021

  24. [24]

    A physics-informed deep learning paradigm for traffic state and fundamental diagram estimation,

    R. Shi, Z. Mo, K. Huang, X. Di, and Q. Du, “A physics-informed deep learning paradigm for traffic state and fundamental diagram estimation,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 8, pp. 11 688–11 698, 2022

    work page 2022

  25. [25]

    Macroscopic traffic flow modeling with physics regularized gaussian process: A new insight into machine learning applications in transportation,

    Y . Yuan, Z. Zhang, X. T. Yang, and S. Zhe, “Macroscopic traffic flow modeling with physics regularized gaussian process: A new insight into machine learning applications in transportation,” Transportation Research Part B: Methodological , vol. 146, pp. 88–110, 2021

    work page 2021

  26. [26]

    Traffic flow modeling with gradual physics regularized learning,

    Y . Yuan, Q. Wang, and X. T. Yang, “Traffic flow modeling with gradual physics regularized learning,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 9, pp. 14 649–14 660, 2022

    work page 2022

  27. [27]

    Physics- informed deep learning for traffic state estimation based on the traffic flow model and computational graph method,

    J. Zhang, S. Mao, L. Yang, W. Ma, S. Li, and Z. Gao, “Physics- informed deep learning for traffic state estimation based on the traffic flow model and computational graph method,” Information Fusion, vol. 101, p. 101971, Jan. 2024

    work page 2024

  28. [28]

    Knowledge-data fusion oriented traffic state estimation: A stochastic physics-informed deep learning approach,

    T. Wang, Y . Li, R. Cheng, G. Zou, T. Dantsujic, and D. Ngoduy, “Knowledge-data fusion oriented traffic state estimation: A stochastic physics-informed deep learning approach,” Sept. 2024

    work page 2024

  29. [29]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” Journal of Computational Physics, vol. 378, pp. 686–707, 2019

    work page 2019

  30. [30]

    Shock waves on the highway,

    P. I. Richards, “Shock waves on the highway,” Oper. Res., vol. 4, no. 1, pp. 42–51, 1956

    work page 1956

  31. [31]

    Resurrection of ‘second order’ models of traffic flow,

    A. Aw and M. Rascle, “Resurrection of ‘second order’ models of traffic flow,”SIAM J. Appl. Math. , vol. 60, no. 3, pp. 916–938, 2000

    work page 2000

  32. [32]

    A non-equilibrium traffic model devoid of gas-like behavior,

    H. M. Zhang, “A non-equilibrium traffic model devoid of gas-like behavior,”Transp. Res. B, Methodol., vol. 36, no. 3, pp. 275–290, 2002

    work page 2002

  33. [33]

    Dpm: A novel training method for physics-informed neural networks in extrapolation,

    J. Kim, K. Lee, D. Lee, S. Y . Jhin, and N. Park, “Dpm: A novel training method for physics-informed neural networks in extrapolation,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 35, no. 9, pp. 8146–8154, May 2021

    work page 2021

  34. [34]

    Physics-informed machine learning,

    G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, and L. Yang, “Physics-informed machine learning,”Nature Reviews Physics, vol. 3, no. 6, pp. 422–440, May 2021

    work page 2021

  35. [35]

    Fourier neural operator for learning solutions to macroscopic traffic flow models: Ap- plication to the forward and inverse problems,

    B. T. Thodi, S. V . R. Ambadipudi, and S. E. Jabari, “Fourier neural operator for learning solutions to macroscopic traffic flow models: Ap- plication to the forward and inverse problems,” Transportation Research Part C: Emerging Technologies, vol. 160, p. 104500, Mar. 2024

    work page 2024

  36. [36]

    Tensor Completion Algorithms in Big Data Analytics,

    Q. Song, H. Ge, J. Caverlee, and X. Hu, “Tensor Completion Algorithms in Big Data Analytics,” ACM Trans. Knowl. Discov. Data, vol. 13, no. 1, pp. 6:1–6:48, Jan. 2019

    work page 2019

  37. [37]

    Fast multivariate spatio- temporal analysis via low rank tensor learning,

    M. T. Bahadori, Q. R. Yu, and Y . Liu, “Fast multivariate spatio- temporal analysis via low rank tensor learning,” in Advances in Neural Information Processing Systems, Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, and K. Weinberger, Eds., vol. 27. Curran Associates, Inc., 2014. [Online]. Available: https://proceedings.neurips.cc/paper files/paper/20...

    work page 2014

  38. [38]

    Network-wide traffic flow estimation with insufficient volume detection and crowdsourcing data,

    Z. Zhang, M. Li, X. Lin, and Y . Wang, “Network-wide traffic flow estimation with insufficient volume detection and crowdsourcing data,” Transportation Research Part C: Emerging Technologies , vol. 121, p. 102870, Dec. 2020

    work page 2020

  39. [39]

    Kernelized probabilistic matrix factorization: Exploiting graphs and side information,

    T. Zhou, H. Shan, A. Banerjee, and G. Sapiro, “Kernelized probabilistic matrix factorization: Exploiting graphs and side information,” in Pro- ceedings of the 2012 SIAM international Conference on Data mining . SIAM, 2012, pp. 403–414

    work page 2012

  40. [40]

    Temporal collaborative filtering with bayesian probabilistic tensor fac- torization,

    L. Xiong, X. Chen, T.-K. Huang, J. Schneider, and J. G. Carbonell, “Temporal collaborative filtering with bayesian probabilistic tensor fac- torization,” in Proceedings of the 2010 SIAM international conference on data mining . SIAM, 2010, pp. 211–222

    work page 2010

  41. [41]

    Temporal Regularized Matrix Factorization for High-dimensional Time Series Prediction,

    H.-F. Yu, N. Rao, and I. S. Dhillon, “Temporal Regularized Matrix Factorization for High-dimensional Time Series Prediction,” inAdvances in Neural Information Processing Systems , vol. 29. Curran Associates, Inc., 2016

    work page 2016

  42. [42]

    Generative adversarial networks,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial networks,” Communications of the ACM , vol. 63, no. 11, pp. 139–144, 2020

    work page 2020

  43. [43]

    Graph Embedding and Extensions: A General Framework for Dimensionality Reduction,

    S. Yan, D. Xu, B. Zhang, H.-j. Zhang, Q. Yang, and S. Lin, “Graph Embedding and Extensions: A General Framework for Dimensionality Reduction,” IEEE Transactions on Pattern Analysis and Machine Intel- ligence, vol. 29, no. 1, pp. 40–51, Jan. 2007

    work page 2007

  44. [44]

    Geo- statistical interpolation of object counts collected from multiple strip transects: Ordinary kriging versus finite domain kriging,

    H. Saito, S. A. McKenna, D. Zimmerman, and T. C. Coburn, “Geo- statistical interpolation of object counts collected from multiple strip transects: Ordinary kriging versus finite domain kriging,” Stochastic Environmental Research and Risk Assessment, vol. 19, pp. 71–85, 2005

    work page 2005

  45. [45]

    Wu, Y ., Zhuang, D., Labbe, A., Sun, L., 2021a

    Y . Wu, D. Zhuang, M. Lei, A. Labbe, and L. Sun, “Spatial aggregation and temporal convolution networks for real-time kriging,” arXiv preprint arXiv:2109.12144, 2021

    work page arXiv 2021

  46. [46]

    MoGERNN: An Inductive Traffic Predictor for Unobserved Locations in Dynamic Sensing Networks,

    Q. Zhou, Y . Zhang, M. A. Makridis, A. Kouvelas, Y . Wang, and S. Hu, “MoGERNN: An Inductive Traffic Predictor for Unobserved Locations in Dynamic Sensing Networks,” Jan. 2025

    work page 2025

  47. [47]

    Increase: Inductive graph representation learning for spatio-temporal kriging,

    C. Zheng, X. Fan, C. Wang, J. Qi, C. Chen, and L. Chen, “Increase: Inductive graph representation learning for spatio-temporal kriging,” in Proceedings of the ACM Web Conference 2023 , ser. WWW ’23. New York, NY , USA: Association for Computing Machinery, 2023, p. 673–683

    work page 2023

  48. [48]

    Inductive and adaptive graph convolution networks equipped with constraint task for spatial–temporal traffic data kriging,

    T. Wei, Y . Lin, S. Guo, Y . Lin, Y . Zhao, X. Jin, Z. Wu, and H. Wan, “Inductive and adaptive graph convolution networks equipped with constraint task for spatial–temporal traffic data kriging,” Knowledge- Based Systems, vol. 284, p. 111325, 2024

    work page 2024

  49. [49]

    A conditional diffusion model for probabilistic estimation of traffic states at sensor-free locations,

    D. Lei, M. Xu, and S. Wang, “A conditional diffusion model for probabilistic estimation of traffic states at sensor-free locations,” Trans- portation Research Part C: Emerging Technologies, vol. 166, p. 104798, Sept. 2024

    work page 2024

  50. [50]

    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 (ICLR ’18) , 2018

    work page 2018

  51. [51]

    Interpreting and Unifying Graph Neural Networks with An Optimization Framework,

    M. Zhu, X. Wang, C. Shi, H. Ji, and P. Cui, “Interpreting and Unifying Graph Neural Networks with An Optimization Framework,” in Proceed- ings of the Web Conference 2021, ser. WWW ’21. New York, NY , USA: Association for Computing Machinery, June 2021, pp. 1215–1226

    work page 2021

  52. [52]

    Dirichlet Energy Constrained Learning for Deep Graph Neural Net- works,

    K. Zhou, X. Huang, D. Zha, R. Chen, L. Li, S.-H. Choi, and X. Hu, “Dirichlet Energy Constrained Learning for Deep Graph Neural Net- works,” in Advances in Neural Information Processing Systems , vol. 34. Curran Associates, Inc., 2021, pp. 21 834–21 846

    work page 2021

  53. [53]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in International Conference on Learning Representations, 2017. [Online]. Available: https://openreview.net/ forum?id=SJU4ayYgl

    work page 2017

  54. [54]

    Diffusion Improves Graph Learning,

    J. Gasteiger, S. Weiß enberger, and S. G ¨unnemann, “Diffusion Improves Graph Learning,” in Advances in Neural Information Processing Sys- tems, vol. 32. Curran Associates, Inc., 2019

    work page 2019

  55. [55]

    The heat kernel as the pagerank of a graph,

    F. Chung, “The heat kernel as the pagerank of a graph,” Proceedings of the National Academy of Sciences , vol. 104, no. 50, pp. 19 735–19 740, Dec. 2007

    work page 2007

  56. [56]

    Video-Based Vehicle Trajectory Data Collection,

    V . G. Kovvali, V . Alexiadis, and L. Zhang, “Video-Based Vehicle Trajectory Data Collection,” in Transportation Research Board 86th Annual MeetingTransportation Research Board , no. 07-0528, 2007

    work page 2007

  57. [57]

    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 27th International Joint Conference on Artificial Intelligence (IJCAI), 2018

    work page 2018

  58. [58]

    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, ser. CIKM ’22. New York, NY , USA: Association for Computing Machinery, 2022, p. 4454–4458

    work page 2022

  59. [59]

    Timemixer: Decomposable multiscale mixing for time series forecasting,

    S. Wang, H. Wu, X. Shi, T. Hu, H. Luo, L. Ma, J. Y . Zhang, and J. ZHOU, “Timemixer: Decomposable multiscale mixing for time series forecasting,” in The Twelfth International Conference on Learning Representations , 2024. [Online]. Available: https: //openreview.net/forum?id=7oLshfEIC2 14 Qishen Zhou received the B.S. degree in transporta- tion engineerin...

    work page 2024