Ensemble of Quantum Circuit Learning for Knowledge Graph Completion

Abstract

Knowledge Graph (KG) is a graph-structured knowledge base that can be used to provide factual information to AI and its applications. Since KGs often have missing facts, KG completion is a fundamental problem. By leveraging expressive power of quantum circuits, this paper proposes an ensemble technique of quantum circuit learning for KG embeddings to complete the KG. Our method makes bagging of inference results made by multiple quantum circuits to mitigate over-fitting of each embedding space. Experimental results confirm that the ensemble improves the performance of KG completion compared to both classical embedding methods and not-ensemble quantum embedding methods.