Pith. sign in

REVIEW 3 major objections 6 minor 3 cited by

RelGNN: Composite Message Passing for Relational Deep Learning

T0 review · 3 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 RelGNN contends that collapsing two-hop junction-table paths into single atomic routes makes relational GNNs more accurate, winning 27 of 30 benchmark tasks.

desk verdict RelGNN's atomic-route message passing is a plausible architectural contribution, but the SOTA claim rests on a thin baseline set and no error bars; worth a careful revision, not a desk reject. read the letter →

arxiv 2502.06784 v2 pith:IMSWVAJX submitted 2025-02-10 cs.LG cs.AIcs.DB

Tianlang Chen , Charilaos Kanatsoulis , Jure Leskovec This is my paper · ORCID
classification cs.LGcs.AIcs.DB
keywords relationaldeeplearninggraphneuralnetworksatomicroutescompositemessagepassingmany-to-manyrelationshipsjunctiontablesattentiondatabases
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

RelGNN argues that graphs built from relational databases have a structure standard heterogeneous GNNs mishandle: many-to-many relationships are materialized as junction tables, so information between two entity types travels through an intermediate table in two hops, with the intermediate's own signal re-aggregated along the way. The paper introduces atomic routes—the minimal path between source and destination node types—and a composite message-passing scheme that completes that exchange in a single hop, fusing the intermediate and source representations before attention-weighted aggregation. On the 30 RelBench tasks, the resulting model beats the standard heterogeneous GraphSAGE baseline on 27 tasks and matches it on the rest, with gains up to 25 percent. The load-bearing structural premise is that connector tables function mostly as routers, so their semantic content can be down-weighted. If that premise holds, RelGNN offers a default GNN design for relational entity graphs that removes redundant aggregation and scales with schema size rather than row count.

What carries the argument

The central object is the atomic route: a simple path between node types, derived automatically from the schema, that supports a single-hop interaction between source and destination nodes—an edge when the source table has one foreign key, and a path $(source \rightarrow intermediate \rightarrow destination)$ when a table has multiple foreign keys. It carries the argument by collapsing two-hop message passing into one composite message, $m^{(l+1)}_{(dst,mid,src)} = \mathrm{AGGR}(h^{(l)}_{dst}, \{\mathrm{FUSE}(h^{(l)}_{mid}, h^{(l)}_{src})\})$, with FUSE a learned linear mix of intermediate and source embeddings and AGGR a multi-head attention that uses the destination as query. Because routes are computed at the schema-graph level, where nodes are table types rather than rows, the machinery is independent of database size, and per-route weight matrices keep different routes from entangling their signals during the final summation.

What would settle it

Train RelGNN and a standard two-hop heterogeneous GNN on a database whose junction tables are given predictive attributes, such as making transaction amount the sole signal for a churn outcome, and compare test accuracy; if the composite one-hop model does not at least match the two-hop baseline, the router assumption that connector tables carry little semantic content is falsified.

Watch

Extended reading notes

Core claim

The paper's claim is that the two-step message flow induced by primary–foreign key links can be reorganized into one-step routes. For a table with one foreign key, an atomic route is the edge itself; for a table with multiple foreign keys, each pair of referenced tables forms a route (source → intermediate → destination). RelGNN replaces the two-hop aggregation with a single composite step $m^{(l+1)}_{(dst,mid,src)} = \mathrm{AGGR}(h^{(l)}_{dst}, \{\mathrm{FUSE}(h^{(l)}_{mid}, h^{(l)}_{src})\})$, then sums over all routes ending at the destination. FUSE is a learned linear combination of intermediate and source embeddings, and AGGR is multi-head attention with the destination as query; distinct weights are learned for every atomic route. The authors report that this architecture outperforms the standard heterogeneous GNN on 27 of 30 RelBench tasks and achieves parity on the remaining three, and their ablations attribute the gain primarily to the atomic-route formulation rather than to attention.

Load-bearing premise

The claim rests on connector tables with two or more foreign keys acting mainly as routers whose own semantic content can be down-weighted; if a junction table's attributes carry strong predictive signal, collapsing its two-hop path into one composite route could discard that signal.

Editorial extensions

If this is right

  • RelGNN surpasses the standard heterogeneous GNN on 27 of 30 RelBench tasks and matches it on the other three, with the largest relative gains on complex primary–foreign key schemas such as rel-f1 and rel-trial.
  • The ablation shows the improvement persists when attention is removed, indicating that the atomic-route composite message passing itself, not the attention mechanism, is the source of most of the gain.
  • On schemas with little junction-table structure, such as rel-amazon and rel-hm, the relative gains shrink to 0–4 percent, which is consistent with the claim that the value comes from fixing bridge and hub message passing.
  • Because atomic routes are derived from table types rather than rows, the approach stays efficient as databases grow; the active connection count is bounded by the neighborhood sampling configuration.
  • Self-loop tables, as in rel-stack, are an orthogonal difficulty; adding a schema-level relative positional encoding yields roughly a 2 percent improvement on the two rel-stack classification tasks.

Reading between the lines

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

  • An untested boundary of the router assumption: if a junction table carries predictive columns of its own, such as a transaction amount or a review text, treating it as a low-content router and folding it into a composite route could discard signal; a benchmark with informative junction attributes would settle this.
  • The atomic-route principle generalizes beyond RelBench: any schema-level star motif around a multi-foreign-key table can be collapsed into pairwise composite routes, so the same argument should transfer to temporal event graphs, supply-chain networks, and medical claims data where junction tables are ubiquitous.
  • A testable extension is to learn per-table weights that decide how much of the intermediate node's own representation survives the FUSE step, letting the model adapt to connector tables that are semantically rich instead of assuming they are pure routers.
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. The paper proposes RELGNN, a GNN architecture for relational entity graphs built from databases. Its core idea is to detect 'bridge' and 'hub' node types (tables with two or more foreign keys), derive 'atomic routes' from the schema, and replace the usual two-hop message passing through these nodes with a single composite aggregation using attention. The method is evaluated on all 30 tasks of the RelBench benchmark, claiming state-of-the-art performance on 27 of 30 tasks with up to 25% improvement over the standard heterogeneous GraphSAGE baseline.

Significance. If the empirical claims were fully supported, RELGNN would be a valuable architectural default for relational deep learning: it is schema-automatic, model-agnostic, and the ablation in Appendix C credibly isolates the benefit of atomic routes from the attention mechanism by using a GraphSAGE-style aggregation variant. The paper also releases code. However, the current experimental evidence only weakly supports the headline claim: results are reported as five-seed means without variance or significance testing, and the comparison set for the 'state-of-the-art' claim is very small (LightGBM plus one heterogeneous GraphSAGE instantiation; GAT and GIN appear only in the ablation). The significance of the contribution will therefore depend on whether these empirical gaps are closed.

major comments (3)
  1. [Sections 5.1-5.3, Tables 1-3] All performance comparisons are reported only as point estimates averaged over five seeds, with no standard deviations, confidence intervals, or significance tests. Several of the claimed gains are small in absolute terms (e.g., Table 1: rel-amazon user-churn +1%, item-churn 0%, rel-stack user-engagement 0%, user-badge 0%; Table 2: rel-stack post-votes 0%; Table 3: rel-hm user-item-purchase 0%). Without uncertainty estimates, these gains cannot be distinguished from run-to-run noise, and the central claim that RELGNN 'outperforms state-of-the-art baselines on the vast majority of tasks' is not established. The authors should report variance, ideally perform paired significance tests across the five seeds, and state how many of the 30 improvements are significant at a conventional level.
  2. [Abstract, Section 5, Section 6] The 'state-of-the-art' claim is not supported by the chosen baselines. Tables 1-3 compare RELGNN only against LightGBM and a single heterogeneous GraphSAGE implementation from the RelBench pipeline; GAT and GIN are shown only in the ablation tables (Tables 5-7). The related work (Section 6) discusses more recent RDL methods such as ContextGNN and LLM-based approaches, but no experimental comparison is made. To justify 'state-of-the-art,' the paper should compare against at least the current leading RelBench entries (including any published leaderboard results) or explicitly limit the claim to the RelBench reference baseline.
  3. [Section 3.1, Eq. (4)] The design premise that bridge and hub nodes 'function mostly as routers' and contribute little semantic content is asserted without empirical support. The composite message passing in Eq. (4) fuses the intermediate node's embedding with the source embedding via a learnable linear combination, but the model is not evaluated on any schema where junction-table attributes carry strong predictive signal, nor is a variant tested that retains the intermediate node's features separately. The authors should either provide evidence for the router assumption or demonstrate robustness to feature-rich connector tables; otherwise the generality of the approach beyond the specific RelBench schemas is uncertain.
minor comments (6)
  1. [Section 4.1] Atomic routes for multiple foreign keys are described as 'hyperedges' (Definition 4.1, case 2), but they are implemented as ordinary aggregated paths; the terminology is confusing and should be clarified or removed.
  2. [Eq. (5)] Equation (5) includes a projection term W_proj h_dst in addition to the attention aggregation, which acts as a residual connection; this design choice is not discussed.
  3. [Tables 1-3] The relative gain reported in Tables 1-3 is never defined; the authors should state how it is computed, especially for MAE where lower is better.
  4. [Section 5] The paper does not report the values of K for the MAP@K recommendation metrics, nor the hyperparameter settings (hidden dimension, number of layers, fanout) used for the reported results; this information is needed for reproducibility.
  5. [Throughout] Several typos: 'Figure 1 and 3' should be 'Figures 1 and 3' (Section 4.1); Table 2 header contains an extra parenthesis '(RELGNN (ours))'; Appendix D says 'of d is the dimension' and should be 'and d is the dimension'; the dataset name appears inconsistently as 'RELBENCH' and 'RelBench' throughout.
  6. [Tables 5-7] RelGNN without attention occasionally outperforms the full RelGNN with attention (e.g., Table 5 item-churn 82.94 vs 82.64; Table 6 driver-position 3.792 vs 3.798; Table 7 user-item-purchase 0.90 vs 0.77). This is not commented on; the authors should acknowledge that the attention mechanism does not always help, which is consistent with their claim that atomic routes are the main contributor.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: RelGNN's central claim rests on a public benchmark evaluation and on architectural construction from database schema, not on fitted inputs or load-bearing self-citations.

full rationale

RelGNN's construction is not derived from its target results. Atomic routes are defined from primary-foreign key relationships (Definition 4.1), and composite message passing is an architectural instantiation (Equations 3-7) whose parameters are trained by standard supervised losses on RelBench; no parameter is fitted to the test metrics and then renamed a prediction. Evaluation uses RelBench, a public benchmark, with the published heterogeneous GraphSAGE and LightGBM pipelines. Although the benchmark and RDL framework were co-authored by one of the present authors (J. Leskovec), these are external public artifacts with fixed preprocessing and publicly available results, so citing them is independent support rather than a self-citation chain. The paper's central structural premise (bridge/hub nodes act mostly as routers) is an explicitly stated design assumption, not a tautology, and it is not imported from a prior uniqueness theorem. Atomic routes are acknowledged to be reminiscent of meta-paths but are distinguished by automatic schema-level construction; this is a design and implementation contribution, not a renaming of the benchmark's own outputs. No equation reduces to its own input by construction. Weaknesses such as limited baseline breadth and absent significance tests would bear on the strength of the empirical claim, but they are not circularity.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

RelGNN introduces no new physical entity and no fitted constant outside ordinary learned weights. The empirical claim rests on the standard assumptions that RelBench is a fair benchmark and that temporal neighbor sampling prevents leakage, plus the paper's own structural assumption that tables with two or more foreign keys are mostly routers. The main omitted detail is the hyperparameter configuration, which is not reported in the main text.

free parameters (1)
  • RelGNN hyperparameters (hidden dimension, number of layers, attention heads, neighbor sampling fanout)
    Architecture sizes and sampling configurations are selected by validation and are not reported in the main text; the central results depend on these choices.
assumptions (3)
  • domain assumption Bridge and hub nodes act mostly as routers with little semantic content.
    Sec 3.1 motivation; if false, atomic-route message passing could lose information carried by junction tables.
  • domain assumption RelBench tasks, temporal splits, and neighbor sampling prevent label leakage and are a fair evaluation environment.
    Sec 2.2 and Sec 5 adopt the pipeline of Fey et al. (2024) and Robinson et al. (2024); the correctness of the benchmark is assumed, not re-verified here.
  • domain assumption Schema graph construction from primary-foreign keys is a sufficient structural description for message passing.
    Sec 2.1-2.2; the method uses only schema-level topology, ignoring attribute semantics beyond node features.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RelGNN: Composite Message Passing for Relational Deep Learning." pith.science (2026). https://pith.science/paper/IMSWVAJX

@misc{pith2026250206784,
  author       = {Pith},
  title        = {Pith review of: RelGNN: Composite Message Passing for Relational Deep Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IMSWVAJX}},
  note         = {Machine review of arXiv:2502.06784}
}
read the original abstract

Predictive tasks on relational databases are critical in real-world applications spanning e-commerce, healthcare, and social media. To address these tasks effectively, Relational Deep Learning (RDL) encodes relational data as graphs, enabling Graph Neural Networks (GNNs) to exploit relational structures for improved predictions. However, existing RDL methods often overlook the intrinsic structural properties of the graphs built from relational databases, leading to modeling inefficiencies, particularly in handling many-to-many relationships. Here we introduce RelGNN, a novel GNN framework specifically designed to leverage the unique structural characteristics of the graphs built from relational databases. At the core of our approach is the introduction of atomic routes, which are simple paths that enable direct single-hop interactions between the source and destination nodes. Building upon these atomic routes, RelGNN designs new composite message passing and graph attention mechanisms that reduce redundancy, highlight key signals, and enhance predictive accuracy. RelGNN is evaluated on 30 diverse real-world tasks from Relbench (Fey et al., 2024), and achieves state-of-the-art performance on the vast majority of tasks, with improvements of up to 25%. Code is available at https://github.com/snap-stanford/RelGNN.

Figures

Figures reproduced from arXiv: 2502.06784 by the authors.

Figure 1
Figure 1. An illustration of the key concepts in our method. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of key concepts of RDL. (a) An example relational database. (b) Relational tables connected by primary-foreign key relations. (c) Relational entity graph built from relational tables. (d) Subgraph sampled with termporal neighbor sampling. Figures from Fey et al. (2024). Each table is a set T = {v1, ..., vnT }, where the elements vi ∈ T are called rows or entities. Each entity v ∈ T has a unique primary … view at source ↗
Figure 3
Figure 3. All the atomic routes derived from rel-f1 dataset. The primary–foreign key relations of rel-f1 is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Primary-foreign key relationships of datasets in R [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Parameter-Free Encoders Remain Viable for RDB Foundation Models

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Trainable RDB encoders cannot robustly exploit neighborhood labels as fixed foundation-model features or feature-importance signals, so simple parameter-free encoders stay near-SOTA.

  2. UniSAGE: Unifying Static and Dynamic Attributes with Hyper-Structure

    cs.CL 2026-05 conditional novelty 6.0 of 10

    UniSAGE jointly models static and dynamic attributes in a single attribute graph and improves entity classification/regression over prior RDL and GNN baselines.

  3. Predictive Query Language: A Domain-Specific Language for Predictive Modeling on Relational Databases

    cs.DB 2026-02 conditional novelty 6.0 of 10

    A SQL-like domain-specific language that declares predictive tasks on relational databases and automatically generates leakage-free training labels, with batch and low-latency implementations.

Reference graph

Works this paper leans on

44 extracted references · 31 canonical work pages · cited by 3 Pith papers

  1. [1]

    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 format.date year duplicate empty "emp...

  2. [2]

    Translating embeddings for modeling multi-relational data

    Bordes, A., Usunier, N., Garcia-Duran, A., Weston, J., and Yakhnenko, O. Translating embeddings for modeling multi-relational data. In Advances in Neural Information Processing Systems (NeurIPS), 2013

    work page 2013

  3. [3]

    and Guestrin, C

    Chen, T. and Guestrin, C. Xgboost: A scalable tree boosting system. In ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), pp.\ 785--794, 2016

    work page 2016

  4. [4]

    Do we really need complicated model architectures for temporal networks? In The Eleventh International Conference on Learning Representations

    Cong, W., Zhang, S., Kang, J., Yuan, B., Wu, H., Zhou, X., Tong, H., and Mahdavi, M. Do we really need complicated model architectures for temporal networks? In The Eleventh International Conference on Learning Representations

    work page

  5. [5]

    Principal neighbourhood aggregation for graph nets

    Corso, G., Cavalleri, L., Beaini, D., Li \`o , P., and Veli c kovi \'c , P. Principal neighbourhood aggregation for graph nets. Advances in neural information processing systems, 33: 0 13260--13271, 2020

    2020

  6. [6]

    Supervised learning on relational databases with graph neural networks

    Cvitkovic, M. Supervised learning on relational databases with graph neural networks. ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019

    work page 2019

  7. [7]

    V., and Swami, A

    Dong, Y., Chawla, N. V., and Swami, A. metapath2vec: Scalable representation learning for heterogeneous networks. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, pp.\ 135--144, 2017

    work page 2017

  8. [8]

    and Lenssen, J

    Fey, M. and Lenssen, J. E. Fast graph representation learning with pytorch geometric. ICLR 2019 (RLGM Workshop), 2019

    work page 2019

Show all 44 references
  1. [9]

    E., Ranjan, R., Robinson, J., Ying, R., You, J., and Leskovec, J

    Fey, M., Hu, W., Huang, K., Lenssen, J. E., Ranjan, R., Robinson, J., Ying, R., You, J., and Leskovec, J. Relational deep learning: Graph representation learning on relational databases. ICML Position Paper, 2024

    2024

  2. [10]

    Magnn: Metapath aggregated graph neural network for heterogeneous graph embedding

    Fu, X., Zhang, J., Meng, Z., and King, I. Magnn: Metapath aggregated graph neural network for heterogeneous graph embedding. In Proceedings of the web conference 2020, pp.\ 2331--2341, 2020

    2020

  3. [11]

    S., Riley, P

    Gilmer, J., Schoenholz, S. S., Riley, P. F., Vinyals, O., and Dahl, G. E. Neural message passing for quantum chemistry. In International Conference on Machine Learning (ICML), pp.\ 1263–1272, 2017

    2017

  4. [12]

    Inductive representation learning on large graphs

    Hamilton, W., Ying, Z., and Leskovec, J. Inductive representation learning on large graphs. In Advances in Neural Information Processing Systems (NeurIPS), 2017

    2017

  5. [13]

    Hanley, J. A. and McNeil, B. J. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology, 148 0 (3): 0 839--843, 1983

    1983

  6. [14]

    X., and Yu, P

    Hu, B., Shi, C., Zhao, W. X., and Yu, P. S. Leveraging meta-path based context for top-n recommendation with a neural co-attention model. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, pp.\ 1531--1540, 2018

    2018

  7. [15]

    Pytorch frame: A modular framework for multi-modal tabular learning

    Hu, W., Yuan, Y., Zhang, Z., Nitta, A., Cao, K., Kocijan, V., Leskovec, J., and Fey, M. Pytorch frame: A modular framework for multi-modal tabular learning. arXiv preprint arXiv:2404.00776, 2024

    2024 arXiv

  8. [16]

    Heterogeneous graph transformer

    Hu, Z., Dong, Y., Wang, K., and Sun, Y. Heterogeneous graph transformer. In Proceedings of the web conference 2020, pp.\ 2704--2710, 2020

    2020

  9. [17]

    On the stability of expressive positional encodings for graphs

    Huang, Y., Lu, W., Robinson, J., Yang, Y., Zhang, M., Jegelka, S., and Li, P. On the stability of expressive positional encodings for graphs. In The Twelfth International Conference on Learning Representations

  10. [18]

    Kaggle Data Science & Machine Learning Survey , 2022

    Kaggle . Kaggle Data Science & Machine Learning Survey , 2022. Available: https://www.kaggle.com/code/paultimothymooney/kaggle-survey-2022-all-results/notebook

    2022

  11. [19]

    I., Choi, E., Jegelka, S., Leskovec, J., and Ribeiro, A

    Kanatsoulis, C. I., Choi, E., Jegelka, S., Leskovec, J., and Ribeiro, A. Learning efficient positional encodings with graph neural networks. arXiv preprint arXiv:2502.01122, 2025

    2025 arXiv

  12. [20]

    M., Goel, R., Eghbali, S., Ramanan, J., Sahota, J., Thakur, S., Wu, S., Smyth, C., Poupart, P., and Brubaker, M

    Kazemi, S. M., Goel, R., Eghbali, S., Ramanan, J., Sahota, J., Thakur, S., Wu, S., Smyth, C., Poupart, P., and Brubaker, M. Time2vec: Learning a vector representation of time. arXiv preprint arXiv:1907.05321, 2019

    1907 arXiv

  13. [21]

    Lightgbm: A highly efficient gradient boosting decision tree

    Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., and Liu, T.-Y. Lightgbm: A highly efficient gradient boosting decision tree. In Advances in Neural Information Processing Systems (NeurIPS), volume 30, 2017

    2017

  14. [22]

    Bpr: Bayesian personalized ranking from implicit feedback

    Rendle, S., Freudenthaler, C., Gantner, Z., and Schmidt-Thieme, L. Bpr: Bayesian personalized ranking from implicit feedback. arXiv preprint arXiv:1205.2618, 2012

    2012 arXiv

  15. [23]

    E., Yuan, Y., Zhang, Z., He, X., and Leskovec, J

    Robinson, J., Ranjan, R., Hu, W., Huang, K., Han, J., Dobles, A., Fey, M., Lenssen, J. E., Yuan, Y., Zhang, Z., He, X., and Leskovec, J. Relbench: A benchmark for deep learning on relational databases. In The Thirty-eight Conference on Neural Information Processing Systems Dat...

    2024

  16. [24]

    N., Bloem, P., van den Berg, R., Titov, I., and Welling, M

    Schlichtkrull, M., Kipf, T. N., Bloem, P., van den Berg, R., Titov, I., and Welling, M. Modeling relational data with graph convolutional networks. In Gangemi, A., Navigli, R., Vidal, M.-E., Hitzler, P., Troncy, R., Hollink, L., Tordai, A., and Alam, M. (eds.), The Semantic We...

    2018

  17. [25]

    M., Han, J., and Peng, J

    Shang, J., Qu, M., Liu, J., Kaplan, L. M., Han, J., and Peng, J. Meta-path guided embedding for similarity search in large-scale heterogeneous information networks. arXiv preprint arXiv:1610.09769, 2016

    2016 arXiv

  18. [26]

    Heterogeneous graph neural networks

    Shi, C. Heterogeneous graph neural networks. In Wu, L., Cui, P., Pei, J., and Zhao, L. (eds.), Graph Neural Networks: Foundations, Frontiers, and Applications, pp.\ 351--369. Springer Singapore, Singapore, 2022

    2022

  19. [27]

    Shi, C., Li, Y., Zhang, J., Sun, Y., and Philip, S. Y. A survey of heterogeneous information network analysis. IEEE Transactions on Knowledge and Data Engineering, 29 0 (1): 0 17--37, 2016

    2016

  20. [28]

    Easing embedding learning by comprehensive transcription of heterogeneous information networks

    Shi, Y., Zhu, Q., Guo, F., Zhang, C., and Han, J. Easing embedding learning by comprehensive transcription of heterogeneous information networks. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp.\ 2190--2199, 2018

    2018

  21. [29]

    Masked label prediction: Unified message passing model for semi-supervised classification

    Shi, Y., Huang, Z., Feng, S., Zhong, H., Wang, W., and Sun, Y. Masked label prediction: Unified message passing model for semi-supervised classification. In Zhou, Z.-H. (ed.), Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21 , pp...

    2021 doi

  22. [30]

    Deep Learning with Relational Logic Representations

    S \' r, G. Deep Learning with Relational Logic Representations. Czech Technical University, 2021

    2021

  23. [31]

    S., and Wu, T

    Sun, Y., Han, J., Yan, X., Yu, P. S., and Wu, T. Pathsim: Meta path-based top-k similarity search in heterogeneous information networks. Proceedings of the VLDB Endowment, 4 0 (11): 0 992--1003, 2011

    2011

  24. [32]

    Higpt: Heterogeneous graph language model

    Tang, J., Yang, Y., Wei, W., Shi, L., Xia, L., Yin, D., and Huang, C. Higpt: Heterogeneous graph language model. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp.\ 2842--2853, 2024

    2024

  25. [33]

    N., Kaiser, L., and Polosukhin, I

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. Attention is all you need. In Advances in Neural Information Processing Systems (NeurIPS), 2017

    2017

  26. [34]

    Graph attention networks

    Veli c kovi \'c , P., Cucurull, G., Casanova, A., Romero, A., Li \`o , P., and Bengio, Y. Graph attention networks. In International Conference on Learning Representations, 2018

    2018

  27. [35]

    Knowledge graph embedding: A survey of approaches and applications

    Wang, Q., Mao, Z., Wang, B., and Guo, L. Knowledge graph embedding: A survey of approaches and applications. IEEE Transactions on Knowledge and Data Engineering, 29 0 (12): 0 2724--2743, 2017

    2017

  28. [36]

    Neural graph collaborative filtering

    Wang, X., He, X., Wang, M., Feng, F., and Chua, T.-S. Neural graph collaborative filtering. In Proceedings of the 42nd international ACM SIGIR conference on Research and development in Information Retrieval, pp.\ 165--174, 2019 a

    2019

  29. [37]

    Wang, X., Ji, H., Shi, C., Wang, B., Ye, Y., Cui, P., and Yu, P. S. Heterogeneous graph attention network. In The world wide web conference, pp.\ 2022--2032, 2019 b

    2022

  30. [38]

    Knowledge graph embedding by translating on hyperplanes

    Wang, Z., Zhang, J., Feng, J., and Chen, Z. Knowledge graph embedding by translating on hyperplanes. In Proceedings of the AAAI conference on artificial intelligence, volume 28, 2014

    2014

  31. [39]

    Tackling prediction tasks in relational databases with llms

    Wydmuch, M., Borchmann, ., and Grali \'n ski, F. Tackling prediction tasks in relational databases with llms. arXiv preprint arXiv:2411.11829, 2024

    2024 arXiv

  32. [40]

    Representation learning on graphs with jumping knowledge networks

    Xu, K., Li, C., Tian, Y., Sonobe, T., Kawarabayashi, K.-i., and Jegelka, S. Representation learning on graphs with jumping knowledge networks. In International conference on machine learning, pp.\ 5453--5462. PMLR, 2018

    2018

  33. [41]

    How powerful are graph neural networks? In International Conference on Learning Representations (ICLR), 2019

    Xu, K., Hu, W., Leskovec, J., and Jegelka, S. How powerful are graph neural networks? In International Conference on Learning Representations (ICLR), 2019

    2019

  34. [42]

    M., Ying, R., and Leskovec, J

    You, J., Gomes-Selman, J. M., Ying, R., and Leskovec, J. Identity-aware graph neural networks. In Proceedings of the AAAI conference on artificial intelligence, pp.\ 10737--10745, 2021

    2021

  35. [43]

    Contextgnn: Beyond two-tower recommendation systems

    Yuan, Y., Zhang, Z., He, X., Nitta, A., Hu, W., Wang, D., Shah, M., Huang, S., Stojanovi c , B., Krumholz, A., et al. Contextgnn: Beyond two-tower recommendation systems. arXiv preprint arXiv:2411.19513, 2024

    2024 arXiv

  36. [44]

    A deep learning blueprint for relational databases

    Zahradn \' k, L., Neumann, J., and S \' r, G. A deep learning blueprint for relational databases. In NeurIPS 2023 Second Table Representation Learning Workshop, 2023

    2023

Pith tools

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