Interaction content aware network embedding via co-embedding of nodes and edges

Abstract

Network embedding has been a hot topic as it can learn node representations that encode the network structure resulting from node interactions. In this paper, besides the network structure, the interaction content within which each interaction arises is also embedded because it reveals interaction preferences of the two nodes involved. Specifically, we propose interaction content aware network embedding (ICANE) via co-embedding of nodes and edges. The embedding of edges is to learn edge representations that preserve the interaction content, which then can be incorporated into node representations through edge representations. Experiments demonstrate ICANE outperforms five recent network embedding models in visualization, link prediction and classification.