Trust from the past: Bayesian Personalized Ranking based Link Prediction in Knowledge Graphs

• DOI: 10.48550/arxiv.1601.03778
• Publication Date: 2016-01-01
• Authors: Baichuan Zhang; Sutanay Choudhury; Mohammad Al Hasan; Xia Ning; Khushbu Agarwal; Sumit Purohit; Paola Gabriela Pe...
• Keywords: Predicate (mathematical logic); Knowledge graph; Link (geometry)

Link prediction, or predicting the likelihood of a link in knowledge graph based on its existing state is key research task. It differs from traditional prediction task that links are categorized into different predicates and performance generally varies widely. In this work, we propose latent feature embedding model which considers for each predicate disjointly. To learn parameters it utilizes Bayesian personalized ranking optimization technique. Experimental results large-scale bases such as YAGO2 show our approach achieves substantially higher than several state-of-art approaches. We also given topological properties induced by edges indicators graph.