Multi-view Heterogeneous Network Embedding

In the real world, the complex and diverse relations among different objects can be described in the form of networks. At the same time, with the emergence and development of network embedding, it has become an effective tool for processing networked data. However, most existing network embedding methods are designed for single-view networks, which have certain limitations in describing and characterizing the network semantics. Therefore, it motivates us to study the problem of multi-view network embedding. In this paper, we propose a Multi-View Embedding method for Heterogeneous Networks, called MVHNE. It mainly focuses on the preservation of the network structure and the semantics, and we do not process them separately, but consider their mutual dependence instead. Specifically, to simplify heterogeneous networks, a semantics-based multi-view generation approach was explored. Then, based on the generated semantic views, our model has two concerns, namely the preservation of single-view semantics and the enhanced view collaboration. With extensive experiments on three real-world datasets, we confirm the validity of considering the interactions between structure and semantics for multi-view network embedding. Experiments further demonstrate that our proposed method outperforms the existing state-of-the-art methods.