Inter-domain Multi-relational Link Prediction

Multi-relational graph is a ubiquitous and important data structure, allowing flexible representation of multiple types interactions relations between entities. Similar to other graph-structured data, link prediction one the most tasks on multi-relational graphs often used for knowledge completion. When related coexist, it great benefit build larger via integrating smaller ones. The integration requires predicting hidden relational connections entities belonged different (inter-domain prediction). However, this poses real challenge existing methods that are exclusively designed same only (intra-domain In study, we propose new approach tackle inter-domain problem by softly aligning entity distributions domains with optimal transport maximum mean discrepancy regularizers. Experiments real-world datasets show regularizer beneficial considerably improves performance baseline methods.