Graph Neural Network leveraging Higher-order Class Label Connectivity for Heterophilous Graphs
Researchers have developed a new classifier called Label Context Classifier (LCC) to improve node classification in heterophilous graphs. Current Graph Neural Networks (GNNs) struggle with these graphs where nodes with different labels are more connected. LCC addresses this by generating label context embeddings through four types of walks, capturing higher-order class label connectivity. When integrated with existing GNNs, LCC demonstrates superior performance over state-of-the-art methods in heterophilous directed graphs. AI
IMPACT Enhances node classification accuracy in heterophilous graphs, potentially improving applications in areas like recommendation systems and social network analysis.