Tucker decomposition-based temporal knowledge graph completion

Knowledge graphs have been demonstrated to be an effective tool for numerous intelligent applications. However, a large amount of valuable knowledge still exists implicitly in the graphs. To enrich existing graphs, recent years witnessed that many algorithms link prediction and embedding designed infer new facts. But most these studies focus on static ignore temporal information which reflects validity knowledge. Developing model completion is increasingly important task. In this paper, we build tensor decomposition inspired by Tucker order-4 tensor. Furthermore, further improve basic performance, provide three kinds methods including cosine similarity, contrastive learning, reconstruction-based incorporate prior into proposed model. Because core contains number parameters model, thus present two regularization schemes avoid over-fitting problem. By combining with our outperforms baselines explicit margin datasets (i.e. ICEWS2014, ICEWS05-15, GDELT).