Self-supervised role learning for graph neural networks

Keyphrases

• Role Learning 100%
• Graph Neural Network 100%
• Structural Role 40%
• Network Motifs 40%
• Popular 20%
• Homogeneous-heterogeneous 20%
• Performance Gain 20%
• Node-based Approach 20%
• Learning Process 20%
• Learning Framework 20%
• Higher-order Structure 20%
• Connectivity Pattern 20%
• Three-state 20%
• Learning Objectives 20%
• Distributional Hypothesis 20%
• Learning Task 20%
• Complex Networks 20%
• Underlying Graph 20%
• Message Passing 20%
• Arbitrary Graphs 20%
• Agnostic 20%
• Homogeneous Network 20%
• K-hop 20%
• Heterogeneous Networks 20%
• Self-supervised Learning 20%
• Network Localization 20%
• Training Curriculum 20%
• Network Proximity 20%
• Local Neighborhood Structure 20%
• Mutual Information Maximization 20%
• Regularized Learning 20%
• Sparse Training 20%

Computer Science

• Graph Neural Network 100%
• Regularization 20%
• Mutual Information 20%
• Local Neighborhood 20%
• Learning Process 20%
• Performance Gain 20%
• Order Structure 20%
• Message Passing 20%
• Learning Framework 20%
• Complex Networks 20%
• Heterogeneous Network 20%
• Neighborhood Structure 20%
• Self-Supervised Learning 20%