Pith. sign in

REVIEW 3 major objections 6 minor 101 references

AirRadar: Inferring Nationwide Air Quality in China with Deep Neural Networks

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read AirRadar, a masked-reconstruction deep network trained on 1,085 Chinese stations, infers PM2.5 at unmonitored locations with 28–45 percent lower MAE than the prior best method STFNN.

desk verdict AirRadar is a plausible, well-engineered architecture with large reported gains, but its random-masking evaluation doesn't yet prove it works in truly unmonitored regions. read the letter →

arxiv 2501.13141 v2 pith:AMJKKS3Y submitted 2025-01-22 cs.LG cs.AI

classification cs.LGcs.AI
keywords airqualityinferencePM2.5spatialinterpolationmaskedautoencoderFourierneuraloperatorcausalheterogeneityChina
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

AirRadar is a deep neural network for inferring real-time air quality at places without monitoring stations, using readings from stations that do exist. The paper claims that on a year of data from 1,085 stations across mainland China, AirRadar reconstructs PM2.5 at randomly masked stations with a mean absolute error of 6.41, 6.79, and 8.11 at 25, 50, and 75 percent masking, versus 11.14, 11.32, and 11.27 for the strongest baseline, STFNN. That is a 28 to 45 percent error reduction, achieved by a masked-reconstruction objective, a local-and-global spatial learner, and a causal module that down-weights confounding regional context. A sympathetic reading is that AirRadar establishes a practical route to nationwide air quality coverage without building thousands of additional stations.

What carries the argument

The load-bearing mechanism is masked feature reconstruction: a shared learnable token stands in for every unobserved station, and the network is trained to fill in the missing readings. Around that token, two learning modules do the work. The Spatial Learning Module captures local correlation with a wind-aligned dartboard projection, pooling neighboring stations into angular-radial regions at 50 and 200 kilometer ranges, and captures global correlation with a Fourier Neural Operator that performs self-attention-like kernel integration in the frequency domain at $O(N E \log N)$ cost. The Causal Learning Module treats spatial context as a confounder $C$ on the path $X \leftarrow C \rightarrow Y$ and applies backdoor adjustment by drawing $C$ from a variational distribution and dynamically weighting context-specific encoders. The whole network is optimized with an L1 loss computed only on masked target nodes.

What would settle it

Withhold readings from a held-out set of real remote stations in western China, train only under random masking, and compare AirRadar's MAE on those withheld stations with its MAE on randomly masked stations in the same split; if the remote-station error is much larger, the random-masking evaluation overstates real-world accuracy.

Watch

Extended reading notes

Core claim

The central claim is that air quality in unmonitored parts of China can be inferred from observed stations with substantially lower error than existing methods, at national scale. AirRadar treats unobserved locations as nodes whose features are replaced by a shared learnable mask token, and trains the network to reconstruct those features. Its Spatial Learning Module combines a dartboard-projection local attention that respects wind direction with a fast Fourier transform global mixer that captures long-range pollutant transport at near-linear cost. Its Causal Learning Module formalizes regional context as a confounder in a structural causal model and applies backdoor adjustment with adaptively weighted context encoders. Trained on 2018 data from 1,085 stations, it reports MAE 6.41, 6.79, and 8.11 at 25, 50, and 75 percent missing station ratios, compared with STFNN's 11.14, 11.32, and 11.27.

Load-bearing premise

The evaluation assumes that randomly hiding a fraction of existing stations reproduces inference at locations that truly have no station; in reality, unmonitored regions are geographically clustered, sparser, and contextually different, so the reported MAE gains may not carry over.

Editorial extensions

If this is right

  • A single AirRadar model can produce nationwide PM2.5 estimates at arbitrary locations from current and historical station data, not just at known sites.
  • Because the spatial learner avoids quadratic cost, the approach scales to thousands of stations where full self-attention becomes prohibitive.
  • The causal backdoor adjustment is what lets the model transfer across diverse regional contexts, so error does not collapse when a new region has a different population, industry, or climate profile.
  • The reported 28–45 percent MAE reduction over STFNN suggests masked reconstruction is a better objective for spatial inference than field-and-graph regression.

Reading between the lines

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

  • If random masking is a faithful proxy, the same architecture could be extended to other pollutants and to real-time operational maps, since only station readings and meteorology are needed.
  • A straightforward test of the deployment claim would be to train under random masking, then evaluate on a held-out set of actual remote stations whose readings are temporarily withheld; the error gap between random and geographically clustered masking would reveal how much of the reported gain transfers.
  • The mask-token and Fourier-mixing recipe may transfer to other spatial interpolation problems, such as traffic speed, temperature, or precipitation, where station coverage is sparse and context heterogeneity is strong.
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

3 major / 6 minor

Summary. This paper proposes AirRadar, a deep neural network for nationwide real-time air quality inference in China. The model uses learnable mask tokens to reconstruct missing node features, then applies a Spatial Learning Module (local dartboard projection plus global FFT-based mixing) and a Causal Learning Module that assigns adaptive weights to learned context representations. On a year-long dataset from 1,085 monitoring stations, the paper reports MAE values of 6.41, 6.79, and 8.11 at 25%, 50%, and 75% random masking ratios, respectively, a reduction of roughly 28–44.9% relative to the STFNN baseline. The paper also includes ablation studies, hyperparameter analyses, and a description of a deployed web interface.

Significance. If the reported results transfer to genuinely unmonitored locations, AirRadar would be a practically valuable system: it addresses the scalability problem of nationwide inference with near-linear spatial mixing, and the 28–44.9% MAE reduction over STFNN is substantial. The paper provides a code repository link, and the ablation and hyperparameter studies give useful evidence about which components matter. However, the evaluation protocol relies entirely on random masking of existing stations, which does not directly validate the central deployment claim of inferring air quality in areas with no monitoring stations. The causal module is presented as backdoor adjustment, but the learned latent contexts do not come with identification guarantees, so the causal interpretation is not established. These two weaknesses materially limit confidence in the paper's main claims.

major comments (3)
  1. [Experimental Settings and Eq. (11)] The paper's central claim is inference at locations lacking monitoring stations, but all quantitative evidence uses random masking of existing stations: the text says 'We randomly mask a percentage of node features' and Eq. (11) computes the loss only over masked nodes, which are uniformly chosen from the 1,085 stations. In real unmonitored areas, especially western China, stations are sparse and clustered differently, so every masked node in the experiments still has many observed neighbors in the same dense network. The reported MAE reductions in Table 1 may therefore reflect interpolation within the observed station graph rather than extrapolation to genuinely unmonitored regions. The authors should add a leave-region-out or station-free evaluation, for example holding out contiguous geographic regions or areas with no nearby stations, and report performance stratified by local station density. Without such an experiment, the deployment claim is not supported.
  2. [Causal Learning Module, Eqs. (7)–(9), Figure 5] The paper claims that backdoor adjustment is used to estimate P(Y|do(X)) and to remove the confounding effect of context C. However, C is a learned latent variable obtained from a variational distribution Q(C|X), trained end-to-end on the same labels as the prediction task. For Eq. (7) to be a valid backdoor adjustment, the structural causal model in Figure 5(a), the prior P(C), and the conditional P(Y|X,C) need to be identified from data; the manuscript provides no identifiability conditions, no constraints on the learned C, and no causal discovery step. The implementation in Eq. (9), which combines context-specific encoders with learned weights, is a plausible domain-adaptation mechanism, but it does not by itself implement a causal intervention. The authors should either provide formal assumptions under which the learned C corresponds to the confounder in the backdoor criterion, or reframe the module as a context-weighting mechanism for distribution shift rather than a causal adjustment. This distinction matters because the causal claim is one of the three stated contributions.
  3. [Table 1 and Model Comparison (RQ1)] The headline results in Table 1 are reported without standard deviations, number of runs, or statistical significance tests. With many free hyperparameters (number of contexts C, sparsity threshold λ, hidden size E, number of spatial blocks L, FFT weight blocks K̂) and no released dataset, it is difficult to rule out that the 28.0–44.9% improvements over STFNN are partly due to run-to-run variance or to tuning on a specific random mask. The authors should report mean ± std over at least three to five random seeds (both model initialization and mask selection), and ideally provide the dataset or a formal data descriptor. This is a basic requirement for a claim of state-of-the-art performance in a paper with no public benchmark.
minor comments (6)
  1. [Table 1] The header contains the typo 'Y ear' instead of 'Year'; please correct it.
  2. [Hyperparameter Study (RQ3)] The section title 'Hyperarmeter Study' should be 'Hyperparameter Study'.
  3. [Practicality (RQ4)] In the sentence 'the the associated number signifies', the word 'the' is duplicated; please fix this typo.
  4. [Inference & Optimization] The text says 'the reconstrued data' near Eq. (10); 'reconstrued' should be 'reconstructed'.
  5. [Eq. (6)] The soft-thresholding and shrinkage operation S(·, λ) is used but never formally defined; please specify its functional form so that the sparsity threshold λ is unambiguous.
  6. [Experimental Settings] The paper states that the dataset contains air quality and meteorological data, but it does not list the exact variables, normalization, or preprocessing steps; this information should be provided to support reproducibility, particularly since the code repository alone is not enough to reconstruct the dataset.

Circularity Check

1 steps flagged · score 4.0 of 10

Central MAE comparison is externally benchmarked, but the causal module's backdoor adjustment is self-defined by the network's own latent contexts, making the causal contribution partially circular.

  1. self definitional [Methodology, Causal Learning Module, Eqs. (7)-(9)]
    "To overcome this, we treat C as a latent variable obtained by a variational distribution Q(C|X) ... we denote a context as a 0-1 matrix, i.e., ck = stack[c1_k, c2_k, ..., cU_k ] ... Yl+1 = PK k=1 Ψl(Yl)[k] Φl k(Yl) ... The computation represents a realization of c drawn from the conditional distribution Q(C|X = X)."

    The confounder C is not measured externally; Eq. (8) defines a context as the network's own stack of inference units, and its distribution Q(C|X) is implemented by the same network whose output is Eq. (9). The 'backdoor adjustment' of Eq. (7) is therefore not an intervention on an independent pre-treatment variable but a weighted mixture of the model's own encoders. The causal effect P(Y|do(X)) is re-derived from the network's fitted latent structure, so any gain from the module is indistinguishable from added capacity. The causal contribution reduces to the model's own construction.

full rationale

The headline empirical claim, that AirRadar reduces MAE by about 28-45% relative to STFNN at 25%, 50%, and 75% random-mask ratios, is supported by Table 1 against external baselines KNN, RF, SGNP, STGNP, VAE, GAE, GraphMAE, MCAM, and STFNN under a fixed train/validation/test split. That comparison is self-contained and not circular. The random-masking protocol is a distribution-shift concern, not a circularity: masked stations are uniformly selected from the 1,085 stations, whereas true unmonitored regions are spatially clustered and sparse; this undermines external validity but does not make the derivation reduce to its inputs. Self-citations to Liang et al. 2023 for dartboard projection, Feng et al. 2024 as the STFNN baseline, and Xia et al. 2024 for causal treatment are credits or baseline references, not load-bearing uniqueness theorems. The one self-definitional element is the Causal Learning Module: context C is defined as the network's own latent inference units and learned end-to-end, so the claimed backdoor adjustment is a re-description of the network's computation rather than an intervention on an independent confounder. Because this affects a component claim rather than the externally benchmarked accuracy result, the score is 4 rather than 6 or higher.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The model depends on learned hyperparameters, on standard spatial autocorrelation assumptions, on the correctness of the SCM in Figure 5a, and on the assumption that random masking approximates real unmonitored inference. The main invented entity is the latent context C used in the causal module, which has no external validation.

free parameters (6)
  • Number of contexts C = 4
    Chosen from hyperparameter study at 75% missing ratio (Figure 8a); controls capacity of the causal weighting module.
  • Sparsity threshold lambda = 1e-2
    Selected in the hyperparameter study; used for soft-thresholding in the global FFT mixing operation.
  • Hidden size E = 32
    Chosen as a balance of capacity and computational cost in Figure 8c.
  • Number of spatial blocks L = 2
    Default in Implementation Details; the hyperparameter study suggests L=3 gives similar accuracy.
  • Dartboard local ranges = 50 km and 200 km
    Chosen by design for local spatial neighborhoods; not ablated or tuned experimentally.
  • Number of FFT weight blocks K_hat = 2
    Implementation detail; the global part comprises 2 weight blocks, with no reported sensitivity analysis.
assumptions (5)
  • domain assumption Spatial autocorrelation: nearby stations influence each other and distant sources also contribute.
    Invoked to motivate the Spatial Learning Module; the local and global learners are built on this assumption.
  • ad hoc to paper The DAG in Figure 5a, where C affects both X and Y and no other confounders exist, correctly describes air quality data generation.
    The causal module assumes this graph; the paper does not test alternative graphs or additional unobserved confounders.
  • ad hoc to paper Backdoor adjustment with a learned variational Q(C|X) approximates P(Y|do(X)) in Eq. 7.
    Standard backdoor adjustment requires knowing or identifying C; treating C as a learned latent from the same data is an extra modeling assumption.
  • domain assumption Random masking during training is a valid surrogate for unmonitored target nodes.
    This underlies the whole evaluation; real unmonitored areas are not randomly distributed and may have different context.
  • standard math The Fourier kernel integral form can approximate global self-attention.
    Standard FNO result cited from Li et al. 2020; used for global spatial mixing in Eqs. 4 to 6.
invented entities (1)
  • Latent context C (context types as 0-1 matrices)
    purpose: Represents spatial context as a confounder in the SCM; dynamic weighting is meant to implement backdoor adjustment for spatial heterogeneity.
    C is learned end-to-end on the same training data with no external measurement, no identified definition, and no counterfactual validation, so it is an invented latent construct rather than an observed confounder.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AirRadar: Inferring Nationwide Air Quality in China with Deep Neural Networks." pith.science (2026). https://pith.science/paper/AMJKKS3Y

@misc{pith2026250113141,
  author       = {Pith},
  title        = {Pith review of: AirRadar: Inferring Nationwide Air Quality in China with Deep Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AMJKKS3Y}},
  note         = {Machine review of arXiv:2501.13141}
}
read the original abstract

Monitoring real-time air quality is essential for safeguarding public health and fostering social progress. However, the widespread deployment of air quality monitoring stations is constrained by their significant costs. To address this limitation, we introduce \emph{AirRadar}, a deep neural network designed to accurately infer real-time air quality in locations lacking monitoring stations by utilizing data from existing ones. By leveraging learnable mask tokens, AirRadar reconstructs air quality features in unmonitored regions. Specifically, it operates in two stages: first capturing spatial correlations and then adjusting for distribution shifts. We validate AirRadar's efficacy using a year-long dataset from 1,085 monitoring stations across China, demonstrating its superiority over multiple baselines, even with varying degrees of unobserved data. The source code can be accessed at https://github.com/CityMind-Lab/AirRadar.

Figures

Figures reproduced from arXiv: 2501.13141 by the authors.

Figure 1
Figure 1. Illustration of air quality inference. Webster 1990). With deep learning advances, DL-based model have become prominent for capturing complex spa￾tial correlations among monitoring stations (Cheng et al. 2018; Han, Lu, and Chen 2021; Hu et al. 2023; Cheng et al. 2024b). However, these models are mostly limited to small￾scale tasks at the city level. In this work, our focus shifts to inferring air quality span￾ning t… view at source ↗
Figure 2
Figure 2. The user interface of our system. It can produce multi-granular inference results. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The framework of AirRadar. MLP: multi-layer perception. MSA: multi-head self-attention. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Dartboard projection: For a query node (gray [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 6
Figure 6. Figure 6: An explanation of how AirRadar combines the in [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 5
Figure 5. Figure 5: SCMs of (a) Data generation under Real-world [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 8
Figure 8. Figure 8: Hyperparameter study when missing ratio=75%. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 7
Figure 7. Figure 7: Effects of modules across different missing ratios. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

101 extracted references · 54 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Bahdanau, D.; Cho, K.; and Bengio, Y. 2014. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473

  4. [4]

    Bai, L.; Yao, L.; Li, C.; Wang, X.; and Wang, C. 2020. Adaptive graph convolutional recurrent network for traffic forecasting. Advances in neural information processing systems, 33: 17804--17815

  5. [5]

    Z.; and Koltun, V

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

  6. [6]

    Blu, T.; Th \'e venaz, P.; and Unser, M. 2004. Linear interpolation revitalized. IEEE Transactions on Image Processing, 13(5): 710--719

  7. [7]

    H.; Song, X.; and Shibasaki, R

    Cai, Z.; Jiang, R.; Lian, X.; Yang, C.; Wang, Z.; Fan, Z.; Tsubouchi, K.; Kobayashi, H. H.; Song, X.; and Shibasaki, R. 2023. Forecasting Citywide Crowd Transition Process via Convolutional Recurrent Neural Networks. IEEE Transactions on Mobile Computing

  8. [8]

    Chen, L.; Xu, J.; Wu, B.; Qian, Y.; Du, Z.; Li, Y.; and Zhang, Y. 2021. Group-aware graph neural network for nationwide city air quality forecasting. arXiv preprint arXiv:2108.12238

Show all 101 references
  1. [9]

    Cheng, S.; Peng, P.; and Lu, F. 2020. A lightweight ensemble spatiotemporal interpolation model for geospatial data. International Journal of Geographical Information Science, 34(9): 1849--1872

  2. [10]

    Cheng, S.; Wang, L.; Wang, P.; and Lu, F. 2024 a . An ensemble spatial prediction method considering geospatial heterogeneity. International Journal of Geographical Information Science, 1--25

  3. [11]

    Cheng, S.; Zhang, W.; Luo, P.; Wang, L.; and Lu, F. 2024 b . An explainable spatial interpolation method considering spatial stratified heterogeneity. International Journal of Geographical Information Science, 1--27

  4. [12]

    Cheng, W.; Shen, Y.; Zhu, Y.; and Huang, L. 2018. A neural attention model for urban air quality inference: Learning the weights of monitoring stations. In Proceedings of the AAAI conference on artificial intelligence, volume 32

  5. [13]

    Choi, J.; Choi, H.; Hwang, J.; and Park, N. 2022. Graph neural controlled differential equations for traffic forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, 6367--6374

  6. [14]

    Cover, T. 1968. Estimation by the nearest neighbor rule. IEEE Transactions on Information Theory, 14(1): 50--55

  7. [15]

    H.; Nguyen, D

    Do, T. H.; Nguyen, D. M.; Tsiligianni, E.; Aguirre, A. L.; La Manna, V. P.; Pasveer, F.; Philips, W.; and Deligiannis, N. 2019. Matrix completion with variational graph autoencoders: Application in hyperlocal air quality inference. In ICASSP 2019-2019 IEEE International Confer...

  8. [16]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

  9. [17]

    Fang, Y.; Qin, Y.; Luo, H.; Zhao, F.; and Zheng, K. 2023. STWave+: A Multi-Scale Efficient Spectral Graph Attention Network With Long-Term Trends for Disentangled Traffic Flow Forecasting. IEEE Transactions on Knowledge and Data Engineering

  10. [18]

    Fang, Z.; Long, Q.; Song, G.; and Xie, K. 2021. Spatial-temporal graph ode networks for traffic flow forecasting. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 364--373

  11. [19]

    E.; Schir \`o , R.; Inglessis, M.; Ferdinandi, M.; and Incoronato, F

    Fanizza, C.; De Berardis, B.; Ietto, F.; Soggiu, M. E.; Schir \`o , R.; Inglessis, M.; Ferdinandi, M.; and Incoronato, F. 2018. Analysis of major pollutants and physico-chemical characteristics of PM2. 5 at an urban site in Rome. Science of the total environment, 616: 1457--1468

  12. [20]

    M.; and Elyan, E

    Fawagreh, K.; Gaber, M. M.; and Elyan, E. 2014. Random forests: from early developments to recent advancements. Systems Science & Control Engineering: An Open Access Journal, 2(1): 602--609

  13. [21]

    Feng, Y.; Wang, Q.; Xia, Y.; Huang, J.; Zhong, S.; and Liang, Y. 2024. Spatio-Temporal Field Neural Networks for Air Quality Inference. In Larson, K., ed., Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJCAI-24 , 7260--7268. Interna...

  14. [22]

    Garcia Duran, A.; and Niepert, M. 2017. Learning graph representations with embedding propagation. Advances in neural information processing systems, 30

  15. [23]

    Geng, X.; Li, Y.; Wang, L.; Zhang, L.; Yang, Q.; Ye, J.; and Liu, Y. 2019. Spatiotemporal multi-graph convolution network for ride-hailing demand forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 33, 3656--3663

  16. [24]

    Graves, A. 2013. Generating sequences with recurrent neural networks. arXiv preprint arXiv:1308.0850

  17. [25]

    Guibas, J.; Mardani, M.; Li, Z.; Tao, A.; Anandkumar, A.; and Catanzaro, B. 2021. Adaptive fourier neural operators: Efficient token mixers for transformers. arXiv preprint arXiv:2111.13587

  18. [26]

    Guo, S.; Chen, C.; Wang, J.; Ding, Y.; Liu, Y.; Xu, K.; Yu, Z.; and Zhang, D. 2020. A force-directed approach to seeking route recommendation in ride-on-demand service using multi-source urban data. IEEE Transactions on Mobile Computing, 21(6): 1909--1926

  19. [27]

    Guo, S.; Chen, C.; Wang, J.; Liu, Y.; Xu, K.; and Chiu, D. M. 2018. Dynamic price prediction in ride-on-demand service with multi-source urban data. In Proceedings of the 15th EAI International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services, 412--421

  20. [28]

    Guo, S.; Lin, Y.; Feng, N.; Song, C.; and Wan, H. 2019 a . Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 33, 922--929

  21. [29]

    Guo, S.; Lin, Y.; Li, S.; Chen, Z.; and Wan, H. 2019 b . Deep spatial--temporal 3D convolutional neural networks for traffic data forecasting. IEEE Transactions on Intelligent Transportation Systems, 20(10): 3913--3926

  22. [30]

    Guo, S.; Lin, Y.; Wan, H.; Li, X.; and Cong, G. 2021. Learning dynamics and heterogeneity of spatial-temporal graph data for traffic forecasting. IEEE Transactions on Knowledge and Data Engineering, 34(11): 5415--5428

  23. [31]

    Han, P.; Shang, S.; Sun, A.; Zhao, P.; Zheng, K.; and Zhang, X. 2021. Point-of-interest recommendation with global and local context. IEEE Transactions on Knowledge and Data Engineering, 34(11): 5484--5495

  24. [32]

    Han, Q.; Lu, D.; and Chen, R. 2021. Fine-Grained Air Quality Inference via Multi-Channel Attention Model. In IJCAI, 2512--2518

  25. [33]

    Hao, X.; Chen, W.; Yan, Y.; Zhong, S.; Wang, K.; Wen, Q.; and Liang, Y. 2024. UrbanVLP: A Multi-Granularity Vision-Language Pre-Trained Foundation Model for Urban Indicator Prediction. arXiv preprint arXiv:2403.16831

  26. [34]

    Hasenfratz, D.; Saukh, O.; Walser, C.; Hueglin, C.; Fierz, M.; and Thiele, L. 2014. Pushing the spatio-temporal resolution limit of urban air pollution maps. In 2014 IEEE International Conference on Pervasive Computing and Communications (PerCom), 69--77. IEEE

  27. [35]

    He, K.; Chen, X.; Xie, S.; Li, Y.; Doll \'a r, P.; and Girshick, R. 2022. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 16000--16009

  28. [36]

    He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 770--778

  29. [37]

    E.; and Zemel, R

    Hinton, G. E.; and Zemel, R. 1993. Autoencoders, minimum description length and Helmholtz free energy. Advances in neural information processing systems, 6

  30. [38]

    Hou, Z.; Liu, X.; Cen, Y.; Dong, Y.; Yang, H.; Wang, C.; and Tang, J. 2022. Graphmae: Self-supervised masked graph autoencoders. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 594--604

  31. [39]

    S.; and Chen, L

    Hu, J.; Guo, C.; Yang, B.; Jensen, C. S.; and Chen, L. 2018. Recurrent multi-graph neural networks for travel cost prediction. arXiv preprint arXiv:1811.05157

  32. [40]

    Hu, J.; Liang, Y.; Fan, Z.; Chen, H.; Zheng, Y.; and Zimmermann, R. 2023. Graph Neural Processes for Spatio-Temporal Extrapolation. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 752--763

  33. [41]

    Jiang, R.; Yin, D.; Wang, Z.; Wang, Y.; Deng, J.; Liu, H.; Cai, Z.; Deng, J.; Song, X.; and Shibasaki, R. 2021. Dl-traff: Survey and benchmark of deep learning models for urban traffic prediction. In Proceedings of the 30th ACM international conference on information & knowled...

  34. [42]

    Jin, G.; Liang, Y.; Fang, Y.; Huang, J.; Zhang, J.; and Zheng, Y. 2023. Spatio-temporal graph neural networks for predictive learning in urban computing: A survey. arXiv preprint arXiv:2303.14483

  35. [43]

    D.; Saco, P

    Khosravi, K.; Panahi, M.; Golkarian, A.; Keesstra, S. D.; Saco, P. M.; Bui, D. T.; and Lee, S. 2020. Convolutional neural network approach for spatial prediction of flood hazard at national scale of Iran. Journal of Hydrology, 591: 125552

  36. [44]

    P.; and Ba, J

    Kingma, D. P.; and Ba, J. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980

  37. [45]

    P.; and Welling, M

    Kingma, D. P.; and Welling, M. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114

  38. [46]

    N.; and Welling, M

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

  39. [47]

    N.; and Welling, M

    Kipf, T. N.; and Welling, M. 2016 b . Variational graph auto-encoders. arXiv preprint arXiv:1611.07308

  40. [48]

    Lang, J.; Zhang, Y.; Zhou, Y.; Cheng, S.; Chen, D.; Guo, X.; Chen, S.; Li, X.; Xing, X.; Wang, H.; et al. 2017. Trends of PM2. 5 and chemical composition in Beijing, 2000-2015. Aerosol and Air Quality Research, 17(2): 412--425

  41. [49]

    Li, J.; Chen, S.; Pan, X.; Yuan, Y.; and Shen, H.-B. 2022. Cell clustering for spatial transcriptomics data with graph neural networks. Nature Computational Science, 2(6): 399--408

  42. [50]

    Li, Y.; Yu, R.; Shahabi, C.; and Liu, Y. 2017. Diffusion convolutional recurrent neural network: Data-driven traffic forecasting. arXiv preprint arXiv:1707.01926

  43. [51]

    Li, Z.; Kovachki, N.; Azizzadenesheli, K.; Liu, B.; Bhattacharya, K.; Stuart, A.; and Anandkumar, A. 2020 a . Fourier neural operator for parametric partial differential equations. arXiv preprint arXiv:2010.08895

  44. [52]

    Li, Z.; Kovachki, N.; Azizzadenesheli, K.; Liu, B.; Bhattacharya, K.; Stuart, A.; and Anandkumar, A. 2020 b . Neural operator: Graph kernel network for partial differential equations. arXiv preprint arXiv:2003.03485

  45. [53]

    Liang, Y.; Ke, S.; Zhang, J.; Yi, X.; and Zheng, Y. 2018. Geoman: Multi-level attention networks for geo-sensory time series prediction. In IJCAI, volume 2018, 3428--3434

  46. [54]

    Liang, Y.; Xia, Y.; Ke, S.; Wang, Y.; Wen, Q.; Zhang, J.; Zheng, Y.; and Zimmermann, R. 2023. Airformer: Predicting nationwide air quality in china with transformers. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 14329--14337

  47. [55]

    Liu, A.; Zhang, Y.; Zhang, X.; Liu, G.; Zhang, Y.; Li, Z.; Zhao, L.; Li, Q.; and Zhou, X. 2020. Representation learning with multi-level attention for activity trajectory similarity computation. IEEE Transactions on Knowledge and Data Engineering, 34(5): 2387--2400

  48. [56]

    Luo, K.; Zhu, Y.; Chen, W.; Wang, K.; Zhou, Z.; Ruan, S.; and Liang, Y. 2024. Towards Robust Trajectory Representations: Isolating Environmental Confounders with Causal Learning. In Larson, K., ed., Proceedings of the Thirty-Third International Joint Conference on Artificial I...

  49. [57]

    Y.; Zhang, P.; and Zhang, L

    Ma, R.; Liu, N.; Xu, X.; Wang, Y.; Noh, H. Y.; Zhang, P.; and Zhang, L. 2020. Fine-grained air pollution inference with mobile sensing systems: A weather-related deep autoencoder model. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 4(2): 1--21

  50. [58]

    Miao, X.; Wu, Y.; Chen, L.; Gao, Y.; and Yin, J. 2022. An experimental survey of missing data imputation algorithms. IEEE Transactions on Knowledge and Data Engineering

  51. [59]

    A.; and Webster, R

    Oliver, M. A.; and Webster, R. 1990. Kriging: a method of interpolation for geographical information systems. International Journal of Geographical Information System, 4(3): 313--332

  52. [60]

    Pan, Z.; Wang, Z.; Wang, W.; Yu, Y.; Zhang, J.; and Zheng, Y. 2019. Matrix factorization for spatio-temporal neural networks with applications to urban flow prediction. In Proceedings of the 28th ACM international conference on information and knowledge management, 2683--2691

  53. [61]

    Pan, Z.; Zhang, W.; Liang, Y.; Zhang, W.; Yu, Y.; Zhang, J.; and Zheng, Y. 2020. Spatio-temporal meta learning for urban traffic prediction. IEEE Transactions on Knowledge and Data Engineering, 34(3): 1462--1476

  54. [62]

    Pearl, J.; Glymour, M.; and Jewell, N. P. 2016. Causal inference in statistics: A primer. John Wiley & Sons

  55. [63]

    Pearl, J.; et al. 2000. Models, reasoning and inference. Cambridge, UK: CambridgeUniversityPress, 19(2): 3

  56. [64]

    Prabowo, A.; Xue, H.; Shao, W.; Koniusz, P.; and Salim, F. D. 2023. Message passing neural networks for traffic forecasting. arXiv preprint arXiv:2305.05740

  57. [65]

    u l, T.; Sch \

    Rafaj, P.; Kiesewetter, G.; G \"u l, T.; Sch \"o pp, W.; Cofala, J.; Klimont, Z.; Purohit, P.; Heyes, C.; Amann, M.; Borken-Kleefeld, J.; et al. 2018. Outlook for clean air in the context of sustainable development goals. Global Environmental Change, 53: 1--11

  58. [66]

    Salim, F.; and Haque, U. 2015. Urban computing in the wild: A survey on large scale participation and citizen engagement with ubiquitous computing, cyber physical systems, and Internet of Things. International Journal of Human-Computer Studies, 81: 31--48

  59. [67]

    S.; Shad, A.; et al

    Shad, R.; Mesgari, M. S.; Shad, A.; et al. 2009. Predicting air pollution using fuzzy genetic linear membership kriging in GIS. Computers, environment and urban systems, 33(6): 472--481

  60. [68]

    Singh, G.; Yoon, J.; Son, Y.; and Ahn, S. 2019. Sequential neural processes. Advances in Neural Information Processing Systems, 32

  61. [69]

    Sui, Y.; Wang, X.; Wu, J.; Lin, M.; He, X.; and Chua, T.-S. 2022. Causal attention for interpretable and generalizable graph classification. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 1696--1705

  62. [70]

    Sun, J.; Zhang, J.; Li, Q.; Yi, X.; Liang, Y.; and Zheng, Y. 2020. Predicting citywide crowd flows in irregular regions using multi-view graph convolutional networks. IEEE Transactions on Knowledge and Data Engineering, 34(5): 2348--2359

  63. [71]

    Tan, Q.; Liu, N.; Huang, X.; Chen, R.; Choi, S.-H.; and Hu, X. 2022. Mgae: Masked autoencoders for self-supervised learning on graphs. arXiv preprint arXiv:2201.02534

  64. [72]

    Tian, Y.; Dong, K.; Zhang, C.; Zhang, C.; and Chawla, N. V. 2023. Heterogeneous graph masked autoencoders. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 9997--10005

  65. [73]

    Tobler, W. R. 1970. A computer movie simulating urban growth in the Detroit region. Economic geography, 46(sup1): 234--240

  66. [74]

    Vallero, D. A. 2014. Fundamentals of air pollution. Academic press

  67. [75]

    N.; Kaiser, .; and Polosukhin, I

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

  68. [76]

    Wang, C.; Zhu, Y.; Zang, T.; Liu, H.; and Yu, J. 2021 a . Modeling inter-station relationships with attentive temporal graph convolutional network for air quality prediction. In Proceedings of the 14th ACM international conference on web search and data mining, 616--634

  69. [77]

    Wang, H.; Yu, Q.; Liu, Y.; Jin, D.; and Li, Y. 2021 b . Spatio-temporal urban knowledge graph enabled mobility prediction. Proceedings of the ACM on interactive, mobile, wearable and ubiquitous technologies, 5(4): 1--24

  70. [78]

    S.; and Cong, G

    Wang, S.; Bao, Z.; Culpepper, J. S.; and Cong, G. 2021 c . A survey on trajectory data management, analytics, and learning. ACM Computing Surveys (CSUR), 54(2): 1--36

  71. [79]

    Wang, S.; Cao, J.; and Philip, S. Y. 2020. Deep learning for spatio-temporal data mining: A survey. IEEE transactions on knowledge and data engineering, 34(8): 3681--3700

  72. [80]

    S.; and Zhang, Y

    Wang, S.; Qiao, L.; Fang, W.; Jing, G.; Sheng, V. S.; and Zhang, Y. 2022. Air Pollution Prediction Via Graph Attention Network and Gated Recurrent Unit. Computers, Materials & Continua, 73(1)

  73. [81]

    Waters, N. 2017. Tobler’s first law of geography. The international encyclopedia of geography, 1--13

  74. [82]

    Wu, K.; Peng, H.; Chen, M.; Fu, J.; and Chao, H. 2021 a . Rethinking and improving relative position encoding for vision transformer. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 10033--10041

  75. [83]

    Wu, L.; Cui, P.; Pei, J.; Zhao, L.; and Guo, X. 2022. Graph neural networks: foundation, frontiers and applications. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 4840--4841

  76. [84]

    Wu, X.; Huang, C.; Zhang, C.; and Chawla, N. V. 2020. Hierarchically structured transformer networks for fine-grained spatial event forecasting. In Proceedings of the web conference 2020, 2320--2330

  77. [85]

    Wu, Y.; Zhuang, D.; Labbe, A.; and Sun, L. 2021 b . Inductive graph neural networks for spatiotemporal kriging. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 4478--4485

  78. [86]

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

  79. [87]

    Xia, T.; Lin, J.; Li, Y.; Feng, J.; Hui, P.; Sun, F.; Guo, D.; and Jin, D. 2021. 3dgcn: 3-dimensional dynamic graph convolutional network for citywide crowd flow prediction. ACM Transactions on Knowledge Discovery from Data (TKDD), 15(6): 1--21

  80. [88]

    Xia, Y.; Liang, Y.; Wen, H.; Liu, X.; Wang, K.; Zhou, Z.; and Zimmermann, R. 2024. Deciphering spatio-temporal graph forecasting: A causal lens and treatment. Advances in Neural Information Processing Systems, 36

  81. [89]

    Yang, C.; Wu, Q.; Wen, Q.; Zhou, Z.; Sun, L.; and Yan, J. 2022. Towards out-of-distribution sequential event prediction: A causal treatment. Advances in neural information processing systems, 35: 22656--22670

  82. [90]

    Yi, X.; Zhang, J.; Wang, Z.; Li, T.; and Zheng, Y. 2018. Deep distributed fusion network for air quality prediction. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, 965--973

  83. [91]

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

  84. [92]

    Yu, F.; and Koltun, V. 2015. Multi-scale context aggregation by dilated convolutions. arXiv preprint arXiv:1511.07122

  85. [93]

    Zhang, G.; Yue, Y.; Wang, K.; Fang, J.; Sui, Y.; Wang, K.; Liang, Y.; Cheng, D.; Pan, S.; and Chen, T. 2024. Two heads are better than one: Boosting graph sparse training via semantic and topological awareness. arXiv preprint arXiv:2402.01242

  86. [94]

    Zhang, J.; Zheng, Y.; and Qi, D. 2017. Deep spatio-temporal residual networks for citywide crowd flows prediction. In Proceedings of the AAAI conference on artificial intelligence, volume 31

  87. [95]

    Zhang, J.; Zheng, Y.; Qi, D.; Li, R.; and Yi, X. 2016. DNN-based prediction model for spatio-temporal data. In Proceedings of the 24th ACM SIGSPATIAL international conference on advances in geographic information systems, 1--4

  88. [96]

    S.; Xu, J.; Wang, D.; Liu, G.; Zhou, X.; et al

    Zhang, T.; Zhao, P.; Liu, Y.; Sheng, V. S.; Xu, J.; Wang, D.; Liu, G.; Zhou, X.; et al. 2019. Feature-level Deeper Self-Attention Network for Sequential Recommendation. In IJCAI, 4320--4326

  89. [97]

    Zhang, Y.; Li, Y.; Zhou, X.; and Luo, J. 2022. Mest-GAN: Cross-City Urban Traffic Estimation with Me ta S patial-T emporal G enerative A dversarial N etworks. In 2022 IEEE International Conference on Data Mining (ICDM), 733--742. IEEE

  90. [98]

    Zhao, L.; Sun, Q.; Ye, J.; Chen, F.; Lu, C.-T.; and Ramakrishnan, N. 2015. Multi-task learning for spatio-temporal event forecasting. In Proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining, 1503--1512

  91. [99]

    Zheng, C.; Fan, X.; Wang, C.; and Qi, J. 2020. Gman: A graph multi-attention network for traffic prediction. In Proceedings of the AAAI conference on artificial intelligence, volume 34, 1234--1241

  92. [100]

    Zheng, Y.; Liu, F.; and Hsieh, H.-P. 2013. U-air: When urban air quality inference meets big data. In Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining, 1436--1444

  93. [101]

    Zheng, Y.; Yi, X.; Li, M.; Li, R.; Shan, Z.; Chang, E.; and Li, T. 2015. Forecasting fine-grained air quality based on big data. In Proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining, 2267--2276

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.