NEAR: Neighborhood Edge AggregatoR for Graph Classification

Learning graph-structured data with graph neural networks (GNNs) has been recently emerging as an important field because of its wide applicability in bioinformatics, chemoinformatics, social network analysis, and mining. Recent GNN algorithms are based on message passing, which enables GNNs to integrate local structures node features recursively. However, past 1-hop neighborhood passing exposed a risk loss information relationships. In this article, we propose Neighborhood Edge AggregatoR (NEAR), framework that aggregates relations between the nodes via edges. NEAR, can be orthogonally combined Graph Isomorphism Network (GIN), gives integrated describes connected. Therefore, NEAR reflect additional structure each beyond themselves neighborhood. Experimental results multiple classification tasks show our algorithm makes good improvement over other existing GNN-based algorithms.