pith. sign in

arxiv: 2605.24207 · v1 · pith:C6MUN2E2new · submitted 2026-05-22 · 💻 cs.DB · cs.LG

Incorporating Deep Learning Design in Database Queries

Yuval Lev Lubarsky , Dean Light , Boaz Berger , Shunit Agmon , Benny Kimelfeld This is my paper

Pith reviewed 2026-06-30 14:13 UTC · model grok-4.3

classification 💻 cs.DB cs.LG
keywords relational deep learningtuple embeddingsquery liftinggraph neural networksdeclarative machine learningdatabase integration
0
0 comments X

The pith

Database queries can be lifted to jointly handle relational data and learnable tuple embeddings.

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

Deep learning on relational data has required exporting tables to graph formats and running separate neural network code. The paper observes that the way these networks combine embeddings follows the structure of joins and other relational operations. By attaching a learnable embedding vector to each tuple and extending the query operators to transform these vectors, the same computations can be expressed inside the database. This makes relational deep learning declarative and integrated with existing database infrastructure, as shown by an implementation that reproduces several graph models with simple queries.

Core claim

By representing tuple provenance as learnable vector embeddings and lifting relational algebra operators to act on both the data and these embeddings, queries can directly realize the computations performed by graph neural networks over relational data.

What carries the argument

Lifted relational queries that propagate and aggregate tuple embeddings according to the query structure.

If this is right

  • Graph neural network models become expressible as standard database queries.
  • The engineering overhead of data export to external ML systems is eliminated.
  • Database optimizations can be applied directly to neural computations.
  • Models including graph convolutional networks, heterogeneous graph transformers, and hypergraph networks can be implemented this way.

Where Pith is reading between the lines

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

  • The approach may enable training and inference entirely inside the database without data movement.
  • It could generalize to other types of neural architectures that operate on relational structures.
  • Query planners might automatically optimize the embedding computations for better performance.

Load-bearing premise

The interactions induced by relational joins are fully captured by the manipulations that graph neural networks perform on tuple embeddings.

What would settle it

Finding a relational deep learning task where no lifted query reproduces the output of the corresponding graph neural network on the same input data and embeddings.

Figures

Figures reproduced from arXiv: 2605.24207 by Benny Kimelfeld, Boaz Berger, Dean Light, Shunit Agmon, Yuval Lev Lubarsky.

Figure 1
Figure 1. Figure 1: Compilation pipeline of RelaNN. A program is com [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Term graph (logical plan) for DriverAgg. Leaves are database relations and inner nodes are NRA operators. that execute joins, projections, and learned transformations end-to￾end on GPU. Data loaders then connect the corresponding database relations to the physical plan via SQL queries. RelaNN is a Python-embedded DSL parsed with Lark [46]. Trans￾formation operators such as Linear and ReLU are resolved by n… view at source ↗
Figure 4
Figure 4. Figure 4: Equation-to-rule correspondence for the HGT at [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
read the original abstract

Deep learning over relational databases is conventionally realized by translating data into graph representations and applying graph-based neural networks within external frameworks. This round-trip between the database and external machine learning (ML) systems introduces non-trivial engineering overhead. In effect, these graph neural networks operate on tuple embeddings and manipulate them in ways that capture the interactions induced by relational joins. Given this natural correspondence, there is no fundamental reason why specifying a neural network over relational data should be substantially harder than querying it. We propose an approach that naturally integrates deep learning with database queries. The key idea is to associate each tuple with provenance, represented as a vector embedding with learnable parameters. Queries are lifted to operate jointly on data and embeddings, mapping input relations with embedded tuples to output relations with embedded tuples. This approach provides a declarative foundation for relational deep learning, facilitating integration with database systems, optimization, and wide adoption. We describe RelaNN, a proof-of-concept implementation of this approach built on top of PyTorch and cuDF. We illustrate the utility of RelaNN by implementing various graph-learning models, including graph convolutional networks, heterogeneous graph transformers, hypergraph neural networks and deep homomorphism networks. The simplicity of the programs and their competitive runtime performance demonstrate a concrete path toward making the implementation of state-of-the-art neural networks over databases as simple as writing a query.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The paper proposes lifting relational database queries to jointly operate on data tuples and associated learnable provenance embeddings, enabling declarative specification of deep learning models (such as GCNs, heterogeneous graph transformers, and hypergraph NNs) directly over relational data without external graph frameworks. It presents RelaNN, a PyTorch/cuDF proof-of-concept, and claims that the natural correspondence between relational joins and GNN operations on embeddings allows simple query-based implementations with competitive runtime.

Significance. If the lifted operators are shown to be semantically equivalent to reference GNN implementations (including multi-hop aggregation, normalization, and heterogeneous edge handling), the work could meaningfully reduce engineering overhead in relational deep learning and support tighter DB-ML integration. The absence of accuracy results, embedding comparisons, or equivalence verification in the provided description limits the assessed significance to a promising but unvalidated direction.

major comments (2)
  1. [Abstract] Abstract: the central claim that queries can 'faithfully reproduce GNN message passing, aggregation, and update steps' for models like heterogeneous graph transformers rests on an unverified natural correspondence; the manuscript reports only that models 'were implemented' and runtime is competitive, with no accuracy numbers, embedding comparisons, or output-equivalence checks against reference implementations.
  2. [Abstract] The description of RelaNN and the lifted operators provides no derivation, formal semantics, or proof that the embedding manipulations preserve the exact aggregation/normalization behavior of the target GNNs (e.g., attention in heterogeneous transformers or hyperedge aggregation); without this, the declarative foundation claim cannot be evaluated.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for highlighting the need for stronger verification of the claimed correspondence between lifted relational operators and GNN computations. We address the two major comments below and will incorporate revisions to provide the requested evidence and formal details.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that queries can 'faithfully reproduce GNN message passing, aggregation, and update steps' for models like heterogeneous graph transformers rests on an unverified natural correspondence; the manuscript reports only that models 'were implemented' and runtime is competitive, with no accuracy numbers, embedding comparisons, or output-equivalence checks against reference implementations.

    Authors: The manuscript grounds the claim in the natural structural correspondence between relational joins and the multi-hop neighborhood aggregations performed by GNNs, which is illustrated through the concrete RelaNN implementations of GCNs, heterogeneous graph transformers, and hypergraph NNs. We agree, however, that the abstract and evaluation sections would be strengthened by explicit verification. We will add a new subsection reporting (i) output-equivalence checks (element-wise L2 distance and cosine similarity on embeddings) against reference implementations in PyTorch Geometric and (ii) end-to-end accuracy on standard node-classification benchmarks for each model. revision: yes

  2. Referee: [Abstract] The description of RelaNN and the lifted operators provides no derivation, formal semantics, or proof that the embedding manipulations preserve the exact aggregation/normalization behavior of the target GNNs (e.g., attention in heterogeneous transformers or hyperedge aggregation); without this, the declarative foundation claim cannot be evaluated.

    Authors: The current text presents the lifting via an intuitive mapping from join-induced interactions to embedding operations but does not supply a formal semantics or equivalence proof for the more involved cases (attention coefficients, normalization constants, hyperedge pooling). We will revise the manuscript by inserting a new section that (a) defines the lifted relational operators with precise algebraic semantics and (b) sketches the equivalence arguments for the supported GNN families, including the handling of heterogeneous attention and hyperedge aggregation. revision: yes

Circularity Check

0 steps flagged

No circularity: new lifted-query machinery introduced without reduction to fitted inputs or self-citations

full rationale

The paper proposes associating tuples with learnable vector embeddings and lifting relational queries to operate jointly on data and embeddings. This is presented as a new declarative foundation rather than a derivation from prior fitted quantities. No equations define a target quantity in terms of itself, no parameters are fitted on a subset and then renamed as predictions, and no load-bearing self-citations or uniqueness theorems from the authors' prior work are invoked. The implementation of RelaNN and example models (GCN, HGT, hypergraph NNs) serves as direct evidence of utility, keeping the central claim self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central idea rests on the assumption that relational operations can be extended to vector embeddings while preserving the semantics needed for graph neural networks; this is postulated rather than derived from external benchmarks.

free parameters (1)
  • embedding dimension and parameters
    Learnable vector parameters attached to each tuple; dimension and initialization not specified in abstract.
axioms (1)
  • domain assumption Relational joins induce interactions that can be captured by operations on tuple embeddings
    Invoked to justify why lifting queries implements graph neural networks.
invented entities (1)
  • tuple provenance embeddings no independent evidence
    purpose: Represent learnable parameters that capture relational interactions
    New representation introduced to enable the lifted queries.

pith-pipeline@v0.9.1-grok · 5776 in / 1263 out tokens · 34078 ms · 2026-06-30T14:13:53.749592+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Neuro-Relational Programs: Unifying Queries and Neural Computation over Structured Data

    cs.DB 2026-06 unverdicted novelty 6.0

    NRPs extend Datalog with embedding operations to create a single formalism readable as both query plans with trainable parts and neural architectures with relational structure.

Reference graph

Works this paper leans on

60 extracted references · 11 canonical work pages · cited by 1 Pith paper · 2 internal anchors

  1. [1]

    Serge Abiteboul, Marcelo Arenas, Pablo Barceló, Meghyn Bienvenu, Diego Cal- vanese, Claire David, Richard Hull, Eyke Hüllermeier, Benny Kimelfeld, Leonid Libkin, Wim Martens, Tova Milo, Filip Murlak, Frank Neven, Magdalena Ortiz, Thomas Schwentick, Julia Stoyanovich, Jianwen Su, Dan Suciu, Victor Vianu, and Ke Yi. 2018. Research Directions for Principles ...

    2018

  2. [2]

    Bronstein, İsmail İlkan Ceylan, and Matthias Lanzinger

    Linus Bao, Emily Jin, Michael M. Bronstein, İsmail İlkan Ceylan, and Matthias Lanzinger. 2025. Homomorphism Counts as Structural Encodings for Graph Learning. InICLR. OpenReview.net

    2025

  3. [3]

    Thomas Bonald, Nathan de Lara, Quentin Lutz, and Bertrand Charpentier. 2020. Scikit-network: Graph Analysis in Python.Journal of Machine Learning Research 21, 185 (2020), 1–6. http://jmlr.org/papers/v21/20-412.html

    2020

  4. [4]

    Rajesh Bordawekar and Oded Shmueli. 2017. Using Word Embedding to Enable Semantic Queries in Relational Databases. InDEEM@SIGMOD. ACM, 5:1–5:4

    2017

  5. [5]

    Riccardo Cappuzzo, Paolo Papotti, and Saravanan Thirumuruganathan. 2020. Creating Embeddings of Heterogeneous Relational Datasets for Data Integration Tasks. InSIGMOD Conference. ACM, 1335–1349

    2020

  6. [6]

    Lingjiao Chen, Arun Kumar, Jeffrey Naughton, and Jignesh M. Patel. 2017. To- wards Linear Algebra over Normalized Data.Proc. VLDB Endow.10, 11 (2017), 1214–1225

    2017

  7. [7]

    Tianlang Chen, Charilaos Kanatsoulis, and Jure Leskovec. 2025. RelGNN: Com- posite Message Passing for Relational Deep Learning. InICML

    2025

  8. [8]

    Wei-Lin Chiang, Xuanqing Liu, Si Si, Yang Li, Samy Bengio, and Cho-Jui Hsieh

  9. [9]

    InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2019, Anchorage, AK, USA, August 4-8, 2019

    Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2019, Anchorage, AK, USA, August 4-8, 2019. ACM, 257–266. doi:10.1145/3292500.3330925

    work page doi:10.1145/3292500.3330925 2019

  10. [10]

    E. F. Codd. 1970. A Relational Model of Data for Large Shared Data Banks. Commun. ACM13, 6 (1970), 377–387

    1970

  11. [11]

    Tamara Cucumides and Floris Geerts. 2025. From Features to Structure: Task- Aware Graph Construction for Relational and Tabular Learning with GNNs. In Tabular Data Analysis Workshop (TaDA) at VLDB

    2025

  12. [12]

    Alexis Cvetkov-Iliev, Alexandre Allauzen, and Gaël Varoquaux. 2023. Relational data embeddings for feature enrichment with background information.Mach. Learn.112, 2 (2023), 687–720

    2023

  13. [13]

    Pedro Domingos. 2025. Tensor Logic: The Language of AI.arXiv preprint arXiv:2510.12269(2025)

    work page arXiv 2025

  14. [14]

    Matthias Fey. 2019. PyTorch Scatter: Optimized Scatter Operations for Py- Torch. https://github.com/rusty1s/pytorch_scatter. GPU-native scatter_add, scatter_mean, scatter_max with autograd support

    2019

  15. [15]

    Matthias Fey, Weihua Hu, Kexin Huang, Jan Eric Lenssen, Rishabh Ranjan, Joshua Robinson, Rex Ying, Jiaxuan You, and Jure Leskovec. 2024. Position: Relational Deep Learning - Graph Representation Learning on Relational Databases. In ICML. OpenReview.net

    2024

  16. [16]

    Matthias Fey, Weihua Hu, Kexin Huang, Jan Eric Lenssen, Rishabh Ranjan, Joshua Robinson, Rex Ying, Jiaxuan You, and Jure Leskovec. 2024. RelBench: A Benchmark for Deep Learning on Relational Databases. InAdvances in Neural Information Processing Systems 37 (NeurIPS), Datasets and Benchmarks Track. https://arxiv.org/abs/2407.20060

    work page arXiv 2024

  17. [17]

    Matthias Fey and Jan E. Lenssen. 2019. Fast Graph Representation Learning with PyTorch Geometric. InICLR Workshop on Representation Learning on Graphs and Manifolds

    2019

  18. [18]

    Billy Joe Franks, Moshe Eliasof, Semih Cantürk, Guy Wolf, Carola-Bibiane Schön- lieb, Sophie Fellenz, and Marius Kloft. 2025. Towards Graph Foundation Models: A Study on the Generalization of Positional and Structural Encodings.Trans. Mach. Learn. Res.2025 (2025)

    2025

  19. [19]

    Xinyu Fu, Jiani Zhang, Ziqiao Meng, and Irwin King. 2020. MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding. In WWW. ACM / IW3C2, 2331–2341

    2020

  20. [20]

    Boris Glavic. 2021. Data Provenance - Origins, Applications, Algorithms, and Models.Foundations and Trends®in Databases9, 3-4 (2021), 209–441. doi:10. 1561/1900000068

    2021

  21. [21]

    Green, Gregory Karvounarakis, and Val Tannen

    Todd J. Green, Gregory Karvounarakis, and Val Tannen. 2007. Provenance semirings. InPODS. ACM, 31–40

    2007

  22. [22]

    Hamilton, Rex Ying, and Jure Leskovec

    William L. Hamilton, Rex Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs. InProceedings of the 31st International Conference on Neural Information Processing Systems(Long Beach, California, USA)(NIPS’17). Curran Associates Inc., Red Hook, NY, USA, 1025–1035

    2017

  23. [23]

    Hellerstein, Christoper Ré, Florian Schoppmann, Daisy Zhe Wang, Eugene Fratkin, Aleksander Gorajek, Kee Siong Ng, Caleb Welton, Xixuan Feng, Kun Li, and Arun Kumar

    Joseph M. Hellerstein, Christoper Ré, Florian Schoppmann, Daisy Zhe Wang, Eugene Fratkin, Aleksander Gorajek, Kee Siong Ng, Caleb Welton, Xixuan Feng, Kun Li, and Arun Kumar. 2012. The MADlib analytics library: or MAD skills, the SQL.Proc. VLDB Endow.5, 12 (Aug. 2012), 1700–1711. doi:10.14778/2367502. 2367510

    work page doi:10.14778/2367502 2012

  24. [24]

    Ziniu Hu, Yuxiao Dong, Kuansan Wang, and Yizhou Sun. 2020. Heterogeneous Graph Transformer. InProceedings of The Web Conference 2020 (WWW). 2704–

    2020

  25. [25]

    https://arxiv.org/abs/2003.01332

    work page arXiv 2003

  26. [26]

    Valter Hudovernik, Federico López, Vid Kocijan, Akihiro Nitta, Jan Eric Lenssen, Jure Leskovec, and Matthias Fey. 2026. KumoRFM-2: Scaling Foundation Models for Relational Learning.CoRRabs/2604.12596 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  27. [27]

    Hasan M. Jamil. 2024. Toward a Declarative Query Language for Machine Learn- ing. InVLDB Workshops. https://api.semanticscholar.org/CorpusID:273878548

    2024

  28. [28]

    Matthias Jasny, Tobias Ziegler, Tim Kraska, Uwe Roehm, and Carsten Binnig

  29. [29]

    InProceedings of the 2020 ACM SIGMOD International Conference on Management of Data(Portland, OR, USA)(SIGMOD ’20)

    DB4ML - An In-Memory Database Kernel with Machine Learning Support. InProceedings of the 2020 ACM SIGMOD International Conference on Management of Data(Portland, OR, USA)(SIGMOD ’20). Association for Computing Machinery, New York, NY, USA, 159–173. doi:10.1145/3318464.3380575

    work page doi:10.1145/3318464.3380575 2020

  30. [30]

    Fahim Shahriar Khan and Ashraf Aboulnaga. 2025. A Vision for SQL-Based Relational Deep Learning. InVLDB 2025 Workshop: Tabular Data Analysis (TaDA)

    2025

  31. [31]

    Kipf and Max Welling

    Thomas N. Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. InInternational Conference on Learning Representations (ICLR)

    2017

  32. [32]

    Arun Kumar, Jeffrey Naughton, and Jignesh M. Patel. 2015. Learning Generalized Linear Models Over Normalized Data. InSIGMOD Conference. ACM, 1969–1984

    2015

  33. [33]

    Andreas Kunft, Asterios Katsifodimos, Sebastian Schelter, Sebastian Breß, Tilmann Rabl, and Volker Markl. 2019. An Intermediate Representation for Optimizing Machine Learning Pipelines.Proc. VLDB Endow.12, 11 (2019), 1553– 1567

    2019

  34. [34]

    Guoliang Li, Ji Sun, Lijie Xu, Shifu Li, Jiang Wang, and Wen Nie. 2024. Gaussml: An end-to-end in-database machine learning system. In2024 IEEE 40th Interna- tional Conference on Data Engineering (ICDE). IEEE, 5198–5210

    2024

  35. [35]

    Xupeng Li, Bin Cui, Yiru Chen, Wentao Wu, and Ce Zhang. 2017. MLog: Towards Declarative In-Database Machine Learning.Proc. VLDB Endow.10, 12 (2017), 1933–1936

    2017

  36. [36]

    Yuval Lev Lubarsky, Jan Tönshoff, Martin Grohe, and Benny Kimelfeld. 2023. Selecting Walk Schemes for Database Embedding. InCIKM. ACM, 1677–1686

    2023

  37. [37]

    Takanori Maehara and Hoang NT. 2024. Deep Homomorphism Networks. InAdvances in Neural Information Processing Systems 37 (NeurIPS). https://proceedings.neurips.cc/paper_files/paper/2024/file/ 65f54fdf62cd5614dc5715ae7ece4ef6-Paper-Conference.pdf

    2024

  38. [38]

    Haggai Maron, Heli Ben-Hamu, Hadar Serviansky, and Yaron Lipman. 2019. Provably Powerful Graph Networks. InAdvances in Neural Information Processing Systems 32 (NeurIPS)

    2019

  39. [39]

    Sidharth Mudgal, Han Li, Theodoros Rekatsinas, AnHai Doan, Youngchoon Park, Ganesh Krishnan, Rohit Deep, Esteban Arcaute, and Vijay Raghavendra. 2018. Deep Learning for Entity Matching: A Design Space Exploration. InSIGMOD Conference. ACM, 19–34

    2018

  40. [40]

    Luis Müller, Mikhail Galkin, Christopher Morris, and Ladislav Rampásek. 2024. Attending to Graph Transformers.Trans. Mach. Learn. Res.2024 (2024)

    2024

  41. [41]

    NVIDIA. 2024. RAPIDS cuDF: GPU DataFrame Library. https://github.com/ rapidsai/cudf. Pandas-compatible API with GPU acceleration

    2024

  42. [42]

    Dan Olteanu. 2020. The Relational Data Borg is Learning.Proc. VLDB Endow.13, 12 (2020), 3502–3515

    2020

  43. [43]

    Paolo Papotti and Carsten Binnig. 2025. Panel on Neural Relational Data: Tabular Foundation Models, LLMs... or both?Proc. VLDB Endow.18 (2025), 5513–5515. https://api.semanticscholar.org/CorpusID:281247089

    2025

  44. [44]

    Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Alban Desmaison, Luca Antiga, and Adam Lerer

  45. [45]

    https://api.semanticscholar.org/ CorpusID:40027675

    Automatic differentiation in PyTorch. https://api.semanticscholar.org/ CorpusID:40027675

  46. [46]

    Khaled Mohammed Saifuddin, Briana Bumgardner, Farhan Tanvir, and Esra Akbas. 2023. HyGNN: Drug-Drug Interaction Prediction via Hypergraph Neural Network. In2023 IEEE 39th International Conference on Data Engineering (ICDE). 1503–1516. doi:10.1109/ICDE55515.2023.00119

    work page doi:10.1109/icde55515.2023.00119 2023

  47. [47]

    Modeling Relational Data with Graph Convolutional Networks

    Michael Schlichtkrull, Thomas N. Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, and Max Welling. 2018. Modeling Relational Data with Graph Convolutional Networks. InThe Semantic Web – 15th International Confer- ence (ESWC) (Lecture Notes in Computer Science, Vol. 10843). Springer, 593–607. https://arxiv.org/abs/1703.06103

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  48. [48]

    Schüle, Matthias Bungeroth, Alfons Kemper, Stephan Günnemann, and Thomas Neumann

    Maximilian E. Schüle, Matthias Bungeroth, Alfons Kemper, Stephan Günnemann, and Thomas Neumann. 2019. MLearn: A Declarative Machine Learning Language for Database Systems. InDEEM@SIGMOD

    2019

  49. [49]

    Schüle, Matthias Bungeroth, Dimitri Vorona, Alfons Kemper, Stephan Günnemann, and Thomas Neumann

    Maximilian E. Schüle, Matthias Bungeroth, Dimitri Vorona, Alfons Kemper, Stephan Günnemann, and Thomas Neumann. 2019. ML2SQL - Compiling a Declarative Machine Learning Language to SQL and Python. InInternational Conference on Extending Database Technology. https://api.semanticscholar.org/ CorpusID:81990872

    2019

  50. [50]

    Erez Shinan. 2017. Lark: A Parsing Library for Python. https://github.com/lark- parser/lark. Accessed: 2026

    2017

  51. [51]

    Thiviyan Thanapalasingam, Lucas van Berkel, Peter Bloem, and Paul Groth

  52. [52]

    doi:10.7717/peerj-cs.1073

    Relational Graph Convolutional Networks: A Closer Look.PeerJ Computer Science8 (2022), e1073. doi:10.7717/peerj-cs.1073

    work page doi:10.7717/peerj-cs.1073 2022

  53. [53]

    Jan Tönshoff, Neta Friedman, Martin Grohe, and Benny Kimelfeld. 2023. Stable Tuple Embeddings for Dynamic Databases. InICDE. IEEE, 1286–1299. 9

    2023

  54. [54]

    Smola, and Zheng Zhang

    Minjie Wang, Lingfan Yu, Da Zheng, Quan Gan, Yu Gai, Zihao Ye, Mufei Li, Jinjing Zhou, Qi Huang, Chao Ma, Ziyue Huang, Qipeng Guo, Hao Zhang, Haibin Lin, Junbo Zhao, Jinyang Li, Alexander J. Smola, and Zheng Zhang. 2019. Deep Graph Library: Towards Efficient and Scalable Deep Learning on Graphs.ICLR Workshop on Representation Learning on Graphs and Manifo...

    2019

  55. [55]

    Xiao Wang, Houye Ji, Chuan Shi, Bai Wang, Yanfang Ye, Peng Cui, and Philip S. Yu. 2019. Heterogeneous Graph Attention Network. InWWW

    2019

  56. [56]

    Yanbo Wang, Xiyuan Wang, Quan Gan, Minjie Wang, Qibin Yang, David Wipf, and Muhan Zhang. 2025. Griffin: Towards a Graph-Centric Relational Database Foundation Model. InICML (Proceedings of Machine Learning Research). PMLR / OpenReview.net

    2025

  57. [57]

    Yu Philip

    Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and S. Yu Philip. 2020. A comprehensive survey on graph neural networks.IEEE Transactions on Neural Networks and Learning Systems32, 1 (2020), 4–24

    2020

  58. [58]

    Prasanna

    Hanqing Zeng, Hongkuan Zhou, Ajitesh Srivastava, Rajgopal Kannan, and Vik- tor K. Prasanna. 2019. GraphSAINT: Graph Sampling Based Inductive Learning Method.ArXivabs/1907.04931 (2019). https://api.semanticscholar.org/CorpusID: 195886159

    work page arXiv 2019

  59. [59]

    Jianan Zhao, Xiao Wang, Chuan Shi, Binbin Hu, Guojie Song, and Yanfang Ye

  60. [60]

    Heterogeneous Graph Structure Learning for Graph Neural Networks. In AAAI. AAAI Press, 4697–4705. 10