QubitE: Qubit Embedding for Knowledge Graph Completion

Anonymous

Abstract: Knowledge graph embeddings (KGEs) learn low-dimensional representations of entities and relations to predict missing facts based on existing ones.Quantum-based KGEs utilize variational quantum circuits for link prediction and score triples via the probability distribution of measuring the qubit states.But current quantum-based KGEs either lose quantum advantages during optimizing, or require a large number of parameters to store quantum states, thus leading to overfitting and low performance.Besides, they ignore theoretical analysis which are essential for understanding the model performance.To address performance issue and bridge theory gap, we propose QubitE which is lightweight and suitable for link prediction task.In addition, our model preserves quantum advantages which enable quantum logical computing based on semantics.Furthermore, we prove that (1) QubitE is full-expressive; (2) QubitE can infer various relation patterns including symmetry/antisymmetry, inversion, and commutative/non-commutative composition; (3) QubitE subsumes several existing approaches, \eg~DistMult, pRotatE, RotatE, TransE and ComplEx; (4) QubitE owns linear space complexity and linear time complexity.Experiments on multiple benchmark knowledge graphs demonstrate that QubitE can achieve comparable results to the state-of-the-art classical models.

Paper Type: long