Exploiting Path Information for Anchor Based Graph Neural Network

by   Yuheng Lu, et al.

Learning node representation that incorporating information from graph structure benefits wide range of tasks on graph. Majority of existing graph neural networks (GNNs) have limited power in capturing position information for a given node. The idea of positioning nodes with selected anchors has been exploit, yet mainly rely on explicit labeling of distance information. Here we propose Graph Inference Representation (GIR), an anchor based GNN encoding path information related to anchors for each node. Abilities to get position-aware embedding are theoretically and experimentally investigated on GIRs and its core variants. Further, the complementary characteristic of GIRs and typical GNNs embeddings are demonstrated. We show that GIRs get outperformed results on position-aware scenario, and could improve GNNs results by fuse GIRs embedding.


page 1

page 2

page 3

page 4


Position-aware Graph Neural Networks

Learning node embeddings that capture a node's position within the broad...

Dynamic Labeling for Unlabeled Graph Neural Networks

Existing graph neural networks (GNNs) largely rely on node embeddings, w...

Position-aware Structure Learning for Graph Topology-imbalance by Relieving Under-reaching and Over-squashing

Topology-imbalance is a graph-specific imbalance problem caused by the u...

Distance-aware Molecule Graph Attention Network for Drug-Target Binding Affinity Prediction

Accurately predicting the binding affinity between drugs and proteins is...

Distance and Hop-wise Structures Encoding Enhanced Graph Attention Networks

Numerous works have proven that existing neighbor-averaging Graph Neural...

Hierarchical Graph Capsule Network

Graph Neural Networks (GNNs) draw their strength from explicitly modelin...

Structure and position-aware graph neural network for airway labeling

We present a novel graph-based approach for labeling the anatomical bran...