Real Quadratic-Form-Based Graph Pooling for Graph Neural Networks

Graph neural networks (GNNs) have developed rapidly in recent years because they can work over non-Euclidean data and possess promising prediction power many real-word applications. The graph classification problem is one of the central problems networks, aims to predict label a with help training graph-structural datasets. pooling scheme an important part for objective. Previous works typically focus on using linear manner. In this paper, we propose real quadratic-form-based framework classification. quadratic form capture pairwise relationship, which brings stronger expressive than existing forms. Experiments benchmarks verify effectiveness proposed based tasks.