Heterogeneous information network embedding with incomplete multi-view fusion

<正>1 Introduction and main contributions Multi-view based HIN embedding method embeds the multiple views of heterogeneous information networks(HIN) to the low-dimensional space.It can preserve rich structural and semantic information by using HIN data of multiple views[1,2].In real networks,relations are partially observed due to privacy issues or other reasons,resulting in the incompleteness of single view [3,4].Existing methods usually ignore this problem,which lead to data sparse and it is difficult to extract the hidden semantic relationship of HIN.