GNN-Paradigm, Limitation and Future

Abstract

I described the wide range of graph applications, graph learning peculiarities and difficulties. On this basis the existing methods of message passing and the problems they have are described. Finally, I also briefly describe the prospects for geometric diagram learning.