Pith. sign in

REVIEW 4 major objections 6 minor 38 references

Scalable and Robust Physics-Informed Graph Neural Networks for Water Distribution Systems

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

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

Pith's one-line read The paper claims that a physics-informed graph neural network named SPI-GNN can estimate water pressure and flow in city-scale water distribution systems accurately enough to outperform the previous deep-learning surrogate, while scaling…

desk verdict Incremental but genuine extension of PI-GNN that scales to larger WDSs; the headline claims overstate the evidence and the normalization rationale is shaky, but the empirical work deserves review. read the letter →

arxiv 2502.12164 v1 pith:NKQO3QQJ submitted 2025-02-11 cs.NE cs.LGcs.SYeess.SY

Inaam Ashraf , André Artelt , Barbara Hammer This is my paper · ORCID
classification cs.NEcs.LGcs.SYeess.SY
keywords waterdistributionsystemshydraulicstateestimationphysics-informedgraphneuralnetworkssurrogatemodelsout-of-distributionrobustnessdatanormalizationpumpsandpressure-reducingvalvesEPANET
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 is trying to establish that a deep-learning surrogate can replace slow hydraulic simulators for estimating pressure heads and pipe flows in water distribution networks, including realistic networks with pumps and pressure-reducing valves. The proposed model, SPI-GNN, couples a local graph neural network with a global physics-informed algorithm in a two-phase training loop, and uses a normalization scheme that respects the hydraulic equations. If the claims hold, engineers doing planning, expansion, and rehabilitation of water systems could run thousands of what-if simulations in a fraction of the time, on networks larger than previous surrogates could handle, without retraining for every new demand pattern. The paper reports accuracy comparable to or better than the prior PI-GNN model on eleven networks, with stable errors when test demands and pipe diameters move well outside training ranges.

What carries the argument

The load-bearing mechanism is the two-part model: a local learnable GNN ($f_1$) that estimates flows and demands by message passing, and a global physics-informed algorithm ($f_2$) that iteratively reconstructs pressure heads from pipe, pump, and PRV equations and then recomputes flows and demands from conservation of mass. The physics-preserving normalization is the other essential piece: demands are divided by their sum and pipe resistances by $3\sigma_{r'}$, with reservoir heads, PRV settings, and pump parameters rescaled so that the normalized system obeys the same hydraulic laws and de-normalization is exact. Training alternates $T$ iterations of $f_1$ alone with $K-T$ iterations of $f_1$ plus $f_2$, and the loss ties estimated demands, physics-derived demands, and the two flow estimates together.

What would settle it

Compute the ratio of the maximum pipe flow to the sum of demands in a benchmark water distribution network with pumps and loops; if this ratio exceeds 1 by a wide margin in realistic cases, the normalization premise is violated. A direct test: train SPI-GNN on such a network and compare de-normalized heads and flows against the hydraulic simulator on out-of-distribution demands—if the normalized MAE jumps well above the paper's reported levels, the scalability and robustness claims fail for recirculating systems.

Watch

Extended reading notes

Core claim

The central claim is that a physics-informed GNN can perform hydraulic state estimation—recovering nodal pressure heads and link flows from demands, reservoir heads, pipe attributes, and pump and valve settings—accurately and efficiently on water distribution networks of practical size. Three modifications carry the argument: an architecture that replaces SeLU with ReLU and concatenates both directional flow embeddings; a training scheme split into a local GNN phase followed by a global physics-constrained phase that incorporates pumps and pressure-reducing valves; and a normalization that divides demands by their total and pipe resistances by three times their standard deviation, then rescales reservoir heads, PRV settings, and pump parameters so that the hydraulic equations stay intact. On eleven networks ranging from 19 to 935 junctions, the reported mean absolute errors on heads, flows, and demands are below a few percent, the model outperforms the previous state of the art on larger networks, and the error remains below 0.05 normalized MAE even when demand noise rises tenfold.

Load-bearing premise

The method rests on the claim that the largest flow in a water distribution system is close to the sum of all consumer demands, so dividing demands by their sum keeps all flows near one; if looping or pumping creates recirculating flows much larger than total demand, the normalization no longer preserves hydraulics and the model's accuracy and out-of-distribution robustness would degrade.

Editorial extensions

If this is right

  • Engineers can use the surrogate in place of the standard hydraulic simulator for planning, expansion, and rehabilitation studies, cutting the cost of what-if simulations by orders of magnitude; the paper reports speed-ups from roughly 190x to over 21,000x depending on network and demand extremity.
  • The model extends surrogate state estimation to realistic networks with pumps and pressure-reducing valves, which the prior PI-GNN could not handle, and scales to networks of roughly one thousand junctions.
  • Because the surrogate is fully differentiable, it can be repurposed for gradient-based optimization tasks such as optimal sensor placement, pipe rehabilitation decisions, and estimating pressures everywhere from sparse sensor readings.
  • Out-of-distribution robustness means the surrogate can be applied to demand patterns and pipe diameter scenarios not present in training, as long as the normalization premise holds.
  • The two-phase training with a bounded iteration budget avoids the exploding gradients that prevented the prior model from training on larger networks.

Reading between the lines

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

  • The normalization premise that the largest flow is close to the sum of all demands is plausible for tree-like gravity-fed networks but may fail in looped or heavily pumped systems where recirculating flows exceed total demand; testing on such a network would reveal whether the out-of-distribution robustness claim generalizes.
  • The reported out-of-distribution experiments vary demands and pipe diameters but not roughness coefficients, pump curves, or network topology; the normalization argument directly covers only the features it rescales.
  • A temporal extension with tank levels as state variables is needed before the surrogate can handle dynamic simulations, which the paper itself lists as future work.
  • The speed-up figures compare inference against a Python-based hydraulic solver; a comparison against optimized solver backends would give a more conservative efficiency estimate.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes SPI-GNN, a physics-informed graph neural network surrogate for hydraulic state estimation in water distribution systems (WDSs). It extends a prior PI-GNN model by adding pumps and pressure-reducing valves, a two-phase training scheme, a flow-aware message-passing architecture, and a physics-preserving data normalization method. The authors evaluate on eleven WDSs of increasing size, reporting in-distribution accuracy against EPANET/WNTR, robustness to out-of-distribution demands and pipe diameters, and large speed-ups over the hydraulic simulator.

Significance. If the empirical results hold, the paper would deliver a practically relevant surrogate model that is substantially faster than EPANET and more scalable than the prior PI-GNN baseline, with a comparatively large benchmark suite of eleven WDSs and a systematic out-of-distribution protocol. The evaluation against an independent simulator (EPANET/WNTR) and the use of a fixed, non-learned physical correction step are genuine strengths. However, the central justification for the proposed normalization mechanism rests on an unverified premise about maximum flows, and several claims exceed the evidence presented, especially the blanket statement of state-of-the-art superiority and the claim of greater out-of-distribution robustness without a direct baseline comparison.

major comments (4)
  1. [Sections III-D and IV-B] The central justification for the demand-sum normalization is the statement in Section III-D that 'the largest flow in a WDS is close to the sum of all demands in the WDS', and Section IV-B repeats that this restricts the largest flow to around 1 and eradicates exploding gradients. This premise fails for several networks in Table I that have multiple reservoirs or pumps (Anytown, Pescara, Modena, PA1, Balerma, L-Town All Areas). In a pipe connecting two reservoirs at different heads, the flow is driven by the head difference and is independent of consumer demand; pumps can similarly drive recirculating flows that exceed total demand. The normalization algebra is reversible, so de-normalization is not biased, but the claimed gradient-control mechanism is not established. Moreover, because gradient norm clipping is also used in Section IV-B, the paper does not isolate the contribution of the demand-sum normalization to training stability and scalability. Please replace the premise with a provable bound or an empirical gradient-norm analysis, or run an ablation that isolates the normalization from gradient clipping and the two-phase scheme.
  2. [Abstract and Table III] The abstract's blanket claim that the model 'outperforms the current state-of-the-art DL model' is contradicted by the small-network results in Table III. On Hanoi, SPI-GNN has a higher head MAE than PI-GNN (0.013 vs 0.004); on Pescara, heads are 0.052 vs 0.049; on L-Town Area C, demands are 0.126 vs 0.117. The text in Section IV-D1 appropriately says SPI-GNN achieves 'similar MAE' for small WDSs, but the abstract and conclusion do not carry this qualification. The performance claim should either be restricted to larger networks or supported by a consistent superiority measure across all WDSs.
  3. [Section IV-D2 and IV-D3, Fig. 2 and Fig. 3] The out-of-distribution experiments report only SPI-GNN's absolute mean MAE for demands and diameters, with no baseline comparison. The abstract claims the approach is 'more robust to out-of-distribution input features' than the state of the art, but Figures 2 and 3 contain no PI-GNN (or other surrogate) curves under the same protocol. These figures support absolute robustness of SPI-GNN on the tested WDSs, but they do not by themselves support the comparative robustness claim. Please add the baseline under identical OOD conditions, or rephrase the claim as absolute robustness rather than superiority.
  4. [Section III-A2 and Table II] The global physics-informed algorithm f2 is presented as an iterative scheme whose number of iterations J 'varies based on how close q_hat are to q*', but in the experiments K is fixed per WDS (Table II). The paper provides no convergence criterion, error bound, or empirical verification that the fixed iteration range yields a solution of Equations (2), (3), (5), and (7). Because this iterative algorithm is load-bearing for the physical consistency of the final predictions, please specify the stopping rule actually used and provide evidence that the fixed K is sufficient for each WDS, or report the sensitivity of the results to K.
minor comments (6)
  1. [Section III-A1] The notation in 'ˆqe := ¯qein ∥ (−¯qein)' mixes concatenation with scalar negation; please clarify the dimensions of the edge flow representation and how the two directions are assembled.
  2. [Section III-A2] The pump message equation contains typographical errors, including an unmatched parenthesis and an unclear combination of subscripts; please rewrite the formula cleanly and define all variables.
  3. [Section III-D] The scaling factor τ = 1000 in the normalization of pipe resistance is introduced without justification; please explain how it is chosen and whether it affects the results.
  4. [Section IV-A] The data-generation description says diameters are varied 'every 6th sample' while demands are varied across all samples; please clarify the exact sampling procedure for the training, validation, and test splits.
  5. [Table III] The 'Demands (Estimated vs True)' column compares an input quantity (demands) with a model estimate; please explain what this metric measures and why it is informative given that demands are used as inputs.
  6. [Section IV-C] The gradient clipping threshold is set to 0.00001, which is unusually small; please comment on how this threshold was selected and whether the final results are sensitive to it.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: SPI-GNN's accuracy and robustness are evaluated against the external EPANET/WNTR simulator, and its physics-informed module is specified in full rather than imported by citation.

full rationale

SPI-GNN is trained without simulator head or flow labels: the objective in Eq. (10) penalizes demand conservation and consistency between the local GNN f1 and the fixed physics-informed algorithm f2, while all accuracy claims are measured against EPANET/WNTR on held-out samples. The physics-informed algorithm f2 is stated explicitly in Eqs. (6)-(7) and the flow/demand computations, so it is not imported as an unverified black box from the cited prior work. The citation to [8] is to a published AAAI paper by the same group that serves as the baseline being improved, and the architectural and training differences are described locally (ReLU, concatenated bidirectional flow embeddings, two-phase training, pump/PRV handling). The normalization in Sec. III-D is an explicit, reversible transformation whose parameters (sum of demands, three-sigma resistance scale) are computed from each input sample and are not fitted to simulator outputs. The out-of-distribution experiments vary the generative demand and diameter noise and compare the final de-normalized predictions against the simulator, so robustness is not true by construction. The premise that the largest flow is close to the total demand (Secs. III-D and IV-B) may be false for looped or pumped networks and is a scientific correctness risk, but it is not circular: the algebra of the normalization remains reversible and the empirical evaluation is external. No load-bearing step reduces to its own inputs by definition or by self-citation.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim depends mainly on standard hydraulic domain assumptions, plus three ad hoc assumptions: convergence of f2, the max-flow-equals-demand-sum property, and the representativeness of synthetic demand sampling. No new physical entities are introduced.

free parameters (5)
  • Loss weights rho and delta = 0.1, 0.1
    Set by hand to balance demand and flow consistency losses; no sensitivity analysis is reported.
  • Phase-2 iteration range K = Per-WDS ranges, e.g., [5,10] for Anytown to [15,20] for KL
    Chosen per network in Table II; affects training cost and accuracy.
  • Normalization scaling factor tau = 1000
    Introduced in r'_fac to scale resistances; its value appears arbitrary and is not justified.
  • Architecture hyperparameters (latent dim, number of GNN layers) = 128, 5
    Fixed choices used for all WDSs; not varied or justified beyond convention.
  • Gradient clipping norm = 0.00001
    Chosen to control exploding gradients; no analysis of sensitivity is provided.
assumptions (5)
  • domain assumption Hazen-Williams head-loss equations (3)-(4) govern pipe flows.
    Standard EPANET model; the surrogate's f2 and normalization are built on it.
  • domain assumption A unique hydraulic state exists for every input and is computable by EPANET.
    Needed to define ground truth labels and to interpret f2 as recovering the state.
  • ad hoc to paper The iterative max-message-passing algorithm f2 converges to the solution of (2), (3), (5), and (7).
    No convergence proof is given; J is set based on unstated closeness of qhat to q* (Section III-A2).
  • ad hoc to paper The largest flow in a WDS is close to the sum of all demands.
    Used to justify demand-sum normalization and gradient control; not generally true in looped or pumped networks (Sections III-D, IV-B).
  • domain assumption Synthetic demand patterns sampled from N(1,0.1) around base demands are representative of realistic and out-of-distribution conditions.
    The robustness claim is only demonstrated within this synthetic sampling family (Section IV-A).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scalable and Robust Physics-Informed Graph Neural Networks for Water Distribution Systems." pith.science (2026). https://pith.science/paper/NKQO3QQJ

@misc{pith2026250212164,
  author       = {Pith},
  title        = {Pith review of: Scalable and Robust Physics-Informed Graph Neural Networks for Water Distribution Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NKQO3QQJ}},
  note         = {Machine review of arXiv:2502.12164}
}
read the original abstract

Water distribution systems (WDSs) are an important part of critical infrastructure becoming increasingly significant in the face of climate change and urban population growth. We propose a robust and scalable surrogate deep learning (DL) model to enable efficient planning, expansion, and rehabilitation of WDSs. Our approach incorporates an improved graph neural network architecture, an adapted physics-informed algorithm, an innovative training scheme, and a physics-preserving data normalization method. Evaluation results on a number of WDSs demonstrate that our model outperforms the current state-of-the-art DL model. Moreover, our method allows us to scale the model to bigger and more realistic WDSs. Furthermore, our approach makes the model more robust to out-of-distribution input features (demands, pipe diameters). Hence, our proposed method constitutes a significant step towards bridging the simulation-to-real gap in the use of artificial intelligence for WDSs.

Figures

Figures reproduced from arXiv: 2502.12164 by the authors.

Figure 1
Figure 1. Architecture of the model incorporating a two-phase training scheme. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Robustness of SPI-GNN to change in demands. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Robustness of SPI-GNN to change in pipe diameters. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 35 canonical work pages

  1. [8]

    Physics-informed graph neural networks for water distribution systems,

    I. Ashraf, J. Strotherm, L. Hermes, and B. Hammer, “Physics-informed graph neural networks for water distribution systems,” Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 20, pp. 21 905–21 913, Mar. 2024

    work page 2024

  2. [33]

    State estimation in water distribution system via diffusion on the edge space,

    B. Kerimov, M. Yang, R. Taormina, and F. Tscheikner-Gratl, “State estimation in water distribution system via diffusion on the edge space,” Water Research, vol. 274, p. 122980, 2025

    work page 2025

  3. [34]

    Towards transfer- able metamodels for water distribution systems with edge-based graph neural networks,

    B. Kerimov, R. Taormina, and F. Tscheikner-Gratl, “Towards transfer- able metamodels for water distribution systems with edge-based graph neural networks,” Water Research, p. 121933, 2024

    work page 2024

  4. [1]

    Urbanization,

    H. Ritchie and M. Roser, “Urbanization,” Our World in Data , 2018, https://ourworldindata.org/urbanization

    work page 2018

  5. [2]

    Traffic4cast at neurips 2021 - temporal and spatial few-shot transfer learning in gridded geo-spatial processes,

    C. Eichenberger and et al., “Traffic4cast at neurips 2021 - temporal and spatial few-shot transfer learning in gridded geo-spatial processes,” in Proceedings of the NeurIPS 2021 Competitions and Demonstrations Track, vol. 176. PMLR, 2022, pp. 97–112

    work page 2021

  6. [3]

    Artificial intelligence techniques in smart grid: A survey,

    O. A. Omitaomu and H. Niu, “Artificial intelligence techniques in smart grid: A survey,” Smart Cities, vol. 4, no. 2, pp. 548–568, 2021

    work page 2021

  7. [4]

    Leak detection methods in water distribution networks: A comparative survey on artificial intelligence applications,

    M. Kammoun, A. Kammoun, and M. Abid, “Leak detection methods in water distribution networks: A comparative survey on artificial intelligence applications,” Journal of Pipeline Systems Engineering and Practice, vol. 13, no. 3, p. 04022024, 2022

    work page 2022

  8. [5]

    Deep learning for critical infrastructure resilience,

    K. Dick, L. Russell, Y . S. Dosso, F. Kwamena, and J. R. Green, “Deep learning for critical infrastructure resilience,” JIS, vol. 25, no. 2, 2019

    work page 2019

Show all 38 references
  1. [6]

    Challenges, methods, data–a survey of machine learning in water distribution networks,

    V . Vaquet, F. Hinder, A. Artelt, I. Ashraf, J. Strotherm, J. Vaquet, J. Brinkrolf, and B. Hammer, “Challenges, methods, data–a survey of machine learning in water distribution networks,” in ICANN 2024 . Cham: Springer Nature Switzerland, 2024, pp. 155–170

    2024

  2. [7]

    Epanet 2.2 user’s manual, water infrastructure division,

    L. Rossman, H. Woo, M. Tryby, F. Shang, R. Janke, and T. Haxton, “Epanet 2.2 user’s manual, water infrastructure division,” CESER, 2020

    2020

  3. [9]

    A weighting strategy to improve water demand forecasting performance based on spatial correlation between multiple sensors,

    Y . Wu, X. Wang, S. Liu, X. Yu, and X. Wu, “A weighting strategy to improve water demand forecasting performance based on spatial correlation between multiple sensors,” Sustainable Cities and Society , vol. 93, p. 104545, 2023

    2023

  4. [10]

    Machine learning model and strategy for fast and accurate detection of leaks in water supply network,

    X. Fan, X. Zhang, and X. . B. Yu, “Machine learning model and strategy for fast and accurate detection of leaks in water supply network,”Journal of Infrastructure Preservation and Resilience , vol. 2, Apr 2021

    2021

  5. [11]

    Investigating the suitability of concept drift detection for detecting leakages in water distribution networks,

    V . Vaquet, F. Hinder, and B. Hammer, “Investigating the suitability of concept drift detection for detecting leakages in water distribution networks,” in ICPRAM, INSTICC. SciTePress, 2024, pp. 296–303

    2024

  6. [12]

    Taking care of our drinking water: Dealing with sensor faults in water distribution networks,

    V . Vaquet, A. Artelt, J. Brinkrolf, and B. Hammer, “Taking care of our drinking water: Dealing with sensor faults in water distribution networks,” in ICANN. Springer Nature Switzerland, 2022, pp. 682–693

    2022

  7. [13]

    Lost in optimiza- tion of water distribution systems: Better call bayes,

    A. Candelieri, A. Ponti, I. Giordani, and F. Archetti, “Lost in optimiza- tion of water distribution systems: Better call bayes,” Water, 2022

    2022

  8. [14]

    A machine- learning approach for monitoring water distribution networks (wdns),

    R. Magini, M. Moretti, M. A. Boniforti, and R. Guercio, “A machine- learning approach for monitoring water distribution networks (wdns),” Sustainability, vol. 15, no. 4, 2023

    2023

  9. [15]

    Spatial graph convolution neural networks for water distribution systems,

    I. Ashraf, L. Hermes, A. Artelt, and B. Hammer, “Spatial graph convolution neural networks for water distribution systems,” inAdvances in Intelligent Data Analysis XXI, B. Cr´emilleux, S. Hess, and S. Nijssen, Eds. Cham: Springer Nature Switzerland, 2023, pp. 29–41

    2023

  10. [16]

    Reconstructing nodal pres- sures in water distribution systems with graph neural networks,

    G. Hajgat ´o, B. Gyires-T ´oth, and G. Pa ´al, “Reconstructing nodal pres- sures in water distribution systems with graph neural networks,” arXiv preprint arXiv:2104.13619, 2021

    2021 arXiv

  11. [17]

    Graph neural networks for pressure estimation in water distribution systems,

    H. Truong, A. Tello, A. Lazovik, and V . Degeler, “Graph neural networks for pressure estimation in water distribution systems,” Water Resources Research, vol. 60, no. 7, 2024

    2024

  12. [18]

    Graph neural networks for state estimation in water distribution systems: Application of supervised and semisupervised learning,

    L. Xing and L. Sela, “Graph neural networks for state estimation in water distribution systems: Application of supervised and semisupervised learning,” Journal of Water Resources Planning and Management, 2022

    2022

  13. [19]

    Spectral networks and locally connected networks on graphs,

    J. Bruna, W. Zaremba, A. D. Szlam, and Y . LeCun, “Spectral networks and locally connected networks on graphs,” CoRR, 2014

    2014

  14. [20]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in ICLR, 2017

    2017

  15. [21]

    Convolutional neural networks on graphs with fast localized spectral filtering,

    M. Defferrard, X. Bresson, and P. Vandergheynst, “Convolutional neural networks on graphs with fast localized spectral filtering,” NIPS, 2016

    2016

  16. [22]

    Deep convolutional networks on graph-structured data,

    M. Henaff, J. Bruna, and Y . LeCun, “Deep convolutional networks on graph-structured data,” arXiv preprint arXiv:1506.05163 , 2015

    2015 arXiv

  17. [23]

    Cayleynets: Graph convolutional neural networks with complex rational spectral filters,

    R. Levie, F. Monti, X. Bresson, and M. M. Bronstein, “Cayleynets: Graph convolutional neural networks with complex rational spectral filters,” IEEE Transactions on Signal Processing , 2018

    2018

  18. [24]

    Adaptive graph convolutional neural networks,

    R. Li, S. Wang, F. Zhu, and J. Huang, “Adaptive graph convolutional neural networks,” in AAAI, 2018

    2018

  19. [25]

    Inductive representation learning on large graphs,

    W. L. Hamilton, R. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” in NIPS, 2017, pp. 1025–1035

    2017

  20. [26]

    Geometric deep learning on graphs and manifolds using mixture model cnns,

    F. Monti, D. Boscaini, J. Masci, E. Rodola, J. Svoboda, and M. M. Bronstein, “Geometric deep learning on graphs and manifolds using mixture model cnns,” in CVPR, 2017, pp. 5115–5124

    2017

  21. [27]

    Large-scale learnable graph convolutional networks,

    H. Gao, Z. Wang, and S. Ji, “Large-scale learnable graph convolutional networks,” in SIGKDD, 2018, pp. 1416–1424

    2018

  22. [28]

    Learning convolutional neural networks for graphs,

    M. Niepert, M. Ahmed, and K. Kutzkov, “Learning convolutional neural networks for graphs,” in ICML. PMLR, 2016, pp. 2014–2023

    2016

  23. [29]

    How powerful are graph neural networks?

    K. Xu, W. Hu, J. Leskovec, and S. Jegelka, “How powerful are graph neural networks?” in ICLR, 2019

    2019

  24. [30]

    Graph Attention Networks,

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Li `o, and Y . Bengio, “Graph Attention Networks,”ICLR, 2018, accepted as poster

    2018

  25. [31]

    The graph neural network model,

    F. Scarselli, M. Gori, A. C. Tsoi, M. Hagenbuchner, and G. Monfardini, “The graph neural network model,” IEEE Transactions on Neural Networks, vol. 20, no. 1, pp. 61–80, 2009

    2009

  26. [32]

    Hammer, Learning with recurrent neural networks , ser

    B. Hammer, Learning with recurrent neural networks , ser. Leacture Notes in Control and Information Sciences. Springer, 2000, vol. 254

    2000

  27. [35]

    Opf-hgnn: Generalizable heterogeneous graph neural networks for ac optimal power flow,

    S. Ghamizi, A. Ma, J. Cao, and P. Rodriguez Cortes, “Opf-hgnn: Generalizable heterogeneous graph neural networks for ac optimal power flow,” in PESGM, 2024, pp. 1–5

    2024

  28. [36]

    Epanet- benchmarks

    K. Research and U. o. C. Innovation Center of Excellence, “Epanet- benchmarks.” [Online]. Available: https://github.com/KIOS-Research/ EPANET-Benchmarks/tree/master

  29. [37]

    Waterbenchmarkhub,

    A. Artelt and K. Giese, “Waterbenchmarkhub,” https://github.com/ WaterFutures/WaterBenchmarkHub, 2024

    2024

  30. [38]

    An overview of the water network tool for resilience,

    K. Klise, R. Murray, and T. Haxton, “An overview of the water network tool for resilience,” in WDSA/CCWI Joint Conference , 2018

    2018

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.