• Corpus ID: 239009915

Graph Neural Networks with Learnable Structural and Positional Representations

  title={Graph Neural Networks with Learnable Structural and Positional Representations},
  author={Vijay Prakash Dwivedi and Anh Tuan Luu and Thomas Laurent and Yoshua Bengio and Xavier Bresson},
Graph neural networks (GNNs) have become the standard learning architectures for graphs. GNNs have been applied to numerous domains ranging from quantum chemistry, recommender systems to knowledge graphs and natural language processing. A major issue with arbitrary graphs is the absence of canonical positional information of nodes, which decreases the representation power of GNNs to distinguish e.g. isomorphic nodes and other graph symmetries. An approach to tackle this issue is to introduce… 

Figures and Tables from this paper

Generalized Laplacian Positional Encoding for Graph Representation Learning

This paper draws inspiration from the recent success of Laplacian-based positional encoding and develops a novel family of positional encoding schemes for graphs by generalizing the optimization problem that Laplace embedding to more general dissimilarity functions rather than the 2-norm used in the original formulation.

Geodesic Graph Neural Network for Efficient Graph Representation Learning

An efficient GNN framework called Geodesic GNN (GDGNN) is proposed that requires only one GNN run and injects conditional relationships between nodes into the model without labeling and theoretically proves that GDGNN is more powerful than plain GNNs.

Affinity-Aware Graph Networks

This paper explores the use of affinity measures as features in graph neural networks, in particular measures arising from random walks, including effective resistance, hitting and commute times, and proposes message passing networks based on these features.

Structure-Aware Transformer for Graph Representation Learning

This work proposes the Structure-Aware Transformer, a class of simple and accessible graph Transformers built upon a new self-attention mechanism that systematically improves performance relative to the base GNN model, successfully combining the advantages of GNNs and Transformers.

Attending to Graph Transformers

A taxonomy of graph transformer architectures is derived, bringing some order to this emerging field by probing how well graph transformers can recover various graph properties, how well they can deal with heterophilic graphs, and to what extent they prevent over-squashing.

A Generalization of ViT/MLP-Mixer to Graphs

This work introduces a new class of GNNs, called Graph MLP-Mixer, that holds three key properties: they capture long-range dependency and mitigate the issue of over-squashing, offer better speed and memory efficiency with a complexity linear to the number of nodes and edges, and show high expressivity in terms of graph isomorphism.

Long Range Graph Benchmark

The Long Range Graph Benchmark (LRGB) 1 is presented with 5 graph learning datasets that arguably require LRI reasoning to achieve strong performance in a given task and is suitable for benchmarking and exploration of MP-GNNs and Graph Transformer architectures that are intended to capture LRI.

Transformers over Directed Acyclic Graphs

This paper studies transformers over directed acyclic graphs (DAGs) and proposes architecture adaptations tailored to DAGs: an attention mechanism that is more efficient than the regular quadratic complexity of transformers and at the same time faithfully captures the DAG structure.

Understanding and Extending Subgraph GNNs by Rethinking Their Symmetries

The most prominent form of subgraph methods, which employs node-based subgraph selection policies such as ego-networks or node marking and deletion, is studied and a novel Subgraph GNN dubbed SUN is designed, which theoretically unifies previous architectures while providing better empirical performance on multiple benchmarks.

Neural Graph Databases

LPG2vec enables combining predictive power of the most powerful GNNs with the full scope of information encoded in the LPG model, paving the way for neural graph databases, a class of systems where the vast complexity of maintained data will benefit from modern and future graph machine learning methods.



How Powerful are Graph Neural Networks?

This work characterize the discriminative power of popular GNN variants, such as Graph Convolutional Networks and GraphSAGE, and show that they cannot learn to distinguish certain simple graph structures, and develops a simple architecture that is provably the most expressive among the class of GNNs.

Improving Graph Neural Network Expressivity via Subgraph Isomorphism Counting

The expressive power of the proposed Graph Substructure Networks (GSN), a topologically-aware message passing scheme based on substructure encoding, is theoretically analysed, showing that it is strictly more expressive than the WL test, and provide sufficient conditions for universality.

Distance Encoding: Design Provably More Powerful Neural Networks for Graph Representation Learning

A general class of structure-related features, termed Distance Encoding (DE), that assists GNNs in representing any set of nodes, while providing strictly more expressive power than the 1-WL test and significantly outperform other state-of-the-art methods especially designed for the above tasks.

Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks

It is shown that GNNs have the same expressiveness as the Weisfeiler-Leman graph isomorphism heuristic in terms of distinguishing non-isomorphic (sub-)graphs, and a generalization of GNN's is proposed, so-called $k$-dimensional GNNS ($k-GNNs), which can take higher-order graph structures at multiple scales into account.

A Generalization of Transformer Networks to Graphs

A graph transformer with four new properties compared to the standard model, which closes the gap between the original transformer, which was designed for the limited case of line graphs, and graph neural networks, that can work with arbitrary graphs.

Expressive Power of Invariant and Equivariant Graph Neural Networks

It is proved that the first approximation guarantees for practical GNNs are proved, paving the way for a better understanding of their generalization.

GraphiT: Encoding Graph Structure in Transformers

The GraphiT model encodes structural and positional information by leveraging relative positional encoding strategies in self-attention scores based on positive definite kernels on graphs and enumerating and encoding local sub-structures such as paths of short length.

Parameterized Hypercomplex Graph Neural Networks for Graph Classification

This work develops graph neural networks that leverage the properties of hypercomplex feature transformation and presents empirical evidence that the proposed model incorporates a regularization effect, alleviating the risk of overfitting.

Residual Gated Graph ConvNets

This work reviews existing graph RNN and ConvNet architectures, and proposes natural extension of LSTM and Conv net to graphs with arbitrary size, and designs a set of analytically controlled experiments on two basic graph problems to test the different architectures.