The network embedding task is to represent the node in the network as a
low-dimensional vector while incorporating the topological and structural
information. Most existing approaches solve this problem by factorizing a
proximity matrix, either directly or implicitly. In this work, we