Learning Embedding for Knowledge Graph Completion with Hypernetwork

Link prediction in Knowledge Graph, also called knowledge completion, is a significant problem in graph mining and has many applications for large companies. The more accurate the link prediction results will bring satisfaction, reduce and avoid risks, and commercial benefits. Almost all state-of-the-art models focus on the deep learning approach, especially using convolutional neural networks (CNN). By analysing the strengths and weaknesses of the CNN based models, we proposed a better model to improve the performance of the link prediction task. Specifically, we apply a CNN with specific filters generated through the Hypernetwork architecture. Moreover, we increase the depth of the model more than baseline models to help learn more helpful information. Experimental results show that the proposed model gets better results when compared to CNN-base models.