Investigating graph feature learning becomes essentially important with the
emergence of graph data in many real-world applications. Several graph neural
network approaches are proposed for node feature learning and they generally
follow a neighboring information aggregation scheme to learn node features.
While great performance has been achieved, the weight