graph-structured data are widespread in real-world applications, such as
social networks, recommender systems, knowledge graphs, chemical molecules etc.
Despite the success of Euclidean space for graph-related learning tasks, its
ability to model complex patterns is essentially constra