Learning distributed node representations in networks has been attractingincreasing attention recently due to its effectiveness in a variety ofapplications. Existing approaches usually study networks with a single type ofproximity between nodes, which defines a single view of a network. However, inreality there usually exists multiple types of proximities between nodes,yielding networks with multiple views. This paper studies learning noderepresentations for networks with multiple views, which aims to infer robustnode representations across different views. We propose a multi-viewrepresentation learning approach, which promotes the collaboration of differentviews and lets them vote for the robust representations. During the votingprocess, an attention mechanism is introduced, which enables each node to focuson the most informative views. Experimental results on real-world networks showthat the proposed approach outperforms existing state-of-the-art approaches fornetwork representation learning with a single view and other competitiveapproaches with multiple views.
展开▼
