Weisfeiler and Leman go sparse: Towards scalable higher-order graph embeddings

被引:0
|
作者
Morris, Christopher [1 ]
Rattan, Gaurav [2 ]
Mutzel, Petra [3 ]
机构
[1] Polytech Montreal, CERC Data Sci Real Time Decis Making, Montreal, PQ, Canada
[2] Rhein Westfal TH Aachen, Dept Comp Sci, Aachen, Germany
[3] Univ Bonn, Dept Comp Sci, Bonn, Germany
关键词
SHERALI-ADAMS RELAXATIONS; NEURAL-NETWORKS;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Graph kernels based on the 1-dimensional Weisfeiler-Leman algorithm and corrosponding neural architectures recently emerged as powerful tools for (supervised) learning with graphs. However, due to the purely local nature of the algorithms, they might miss essential patterns in the given data and can only handle binary relations. The k-dimensional Weisfeiler-Leman algorithm addresses this by considering k-tuples, defined over the set of vertices, and defines a suitable notion of adjacency between these vertex tuples. Hence, it accounts for the higher-order interactions between vertices. However, it does not scale and may suffer from overfitting when used in a machine learning setting. Hence, it remains an important open problem to design WL-based graph learning methods that are simultaneously expressive, scalable, and non-overfitting. Here, we propose local variants and corresponding neural architectures, which consider a subset of the original neighborhood, making them more scalable, and less prone to overfitting. The expressive power of (one of) our algorithms is strictly higher than the original algorithm, in terms of ability to distinguish non-isomorphic graphs. Our experimental study confirms that the local algorithms, both kernel and neural architectures, lead to vastly reduced computation times, and prevent overfitting. The kernel version establishes a new state-of-the-art for graph classification on a wide range of benchmark datasets, while the neural version shows promising performance on large-scale molecular regression tasks.
引用
收藏
页数:17
相关论文
共 50 条
  • [31] Scalable Higher-Order Tensor Product Spline Models
    Ruegamer, David
    INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 238, 2024, 238
  • [32] Hintikka's World: Scalable Higher-order Knowledge
    Charrier, Tristan
    Gamblin, Sebastien
    Niveau, Alexandre
    Schwarzentruber, Francois
    PROCEEDINGS OF THE TWENTY-EIGHTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2019, : 6494 - 6496
  • [33] Sparse Learning of Higher-Order Statistics for Communications and Sensing
    Sun, Zhuo
    Kong, Song
    Wang, Wenbo
    IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE, 2020, 4 (01): : 13 - 22
  • [34] Towards a Higher-Order Mathematical Operational Semantics
    Goncharov, Sergey
    Milius, Stefan
    Schroeder, Lutz
    Tsampas, Stelios
    Urbat, Henning
    PROCEEDINGS OF THE ACM ON PROGRAMMING LANGUAGES-PACMPL, 2023, 7 (POPL): : 632 - 658
  • [35] Incorporating Higher-order Structural Information for Graph Clustering
    Li, Qiankun
    Liu, Haobing
    Jiang, Ruobing
    Wang, Tingting
    DATABASE SYSTEMS FOR ADVANCED APPLICATIONS, DASFAA 2024, PT IV, 2024, 14853 : 507 - 517
  • [36] A Graph-Based Higher-Order Intermediate Representation
    Leissa, Roland
    Koester, Marcel
    Hack, Sebastian
    2015 IEEE/ACM INTERNATIONAL SYMPOSIUM ON CODE GENERATION AND OPTIMIZATION (CGO), 2015, : 202 - 212
  • [37] Local structure graph models with higher-order dependence
    Casleton, Emily M.
    Nordman, Daniel J.
    Kaiser, Mark S.
    CANADIAN JOURNAL OF STATISTICS-REVUE CANADIENNE DE STATISTIQUE, 2021, 49 (02): : 497 - 513
  • [38] Higher-order graph wavelets and sparsity on circulant graphs
    Kotzagiannidis, Madeleine S.
    Dragotti, Pier Luigi
    WAVELETS AND SPARSITY XVI, 2015, 9597
  • [39] Higher-order fluctuations in dense random graph models
    Kaur, Gursharn
    Rollin, Adrian
    ELECTRONIC JOURNAL OF PROBABILITY, 2021, 26
  • [40] Graph Representations for Higher-Order Logic and Theorem Proving
    Paliwal, Aditya
    Loos, Sarah M.
    Rabe, Markus N.
    Bansal, Kshitij
    Szegedy, Christian
    THIRTY-FOURTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THE THIRTY-SECOND INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE AND THE TENTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2020, 34 : 2967 - 2974