Inspired by the extensive success of deep learning, graph neural networks
(GNNs) have been proposed to learn expressive node representations and
demonstrated promising performance in various graph learning tasks. However,
existing endeavors predominately focus on the conventional semi-