Multi-relational Propagation (MrP) suggests a propagation approach designed for CONTINUOUS labels, and how to complete them on multi-relational and directed graphs.
In this study, we address the incompleteness in the numerical node attributes of a knowledge graph. We propose the algorithm MrAP, Multi-Relational Attribute Propagation, which we formulate within a message passing scheme.
The structure of the data very often depends on complex relationships of multiple types. In this study, our main research question is "Can we support the structure inference process with a priori relational information about the data domain?"