Efficient Inference of Markov Logic Network for Link Prediction

Zhongbin Li, Liang Duan, Kun Yue

Abstract: Link prediction has achieved great success on ubiquitous graph-based applications, which usually contain multiple types of connections. The heterogeneity of networks introduces complexities in two aspects: representation of multiple types of links, and incorporation of domain knowledge. In this paper, we construct Markov logic network (MLN) to preserve he logical information of graphs for link prediction tasks. By using the graph neural network, we propose a novel inference method in MLNs to guarantee the scalability of link prediction. Experimental results on various datasets show that our proposed method outperforms the state-of-the-art competitors on precision and efficiency.