Researchers have developed a unified framework for Variational Quantum Algorithms (VQAs) applied to knowledge graph embeddings, proposing a new variant that reduces hardware requirements. VQAs combine quantum circuits with classical optimization to address problems that could benefit from current quantum hardware (NISQ). In the context of knowledge graph embeddings, existing proposals differ in their scoring function and the number of qubits needed. This new approach seeks to improve efficiency and interpretability in these systems.

Previous architectures for knowledge graph embeddings in VQAs used two main designs. One employed $n+1$ qubits and obtained the score through a swap test on an auxiliary qubit. The other used $2n+1$ qubits and applied a swap test between two registers. In both cases, entities and relations were represented in a Hilbert space of dimension $d = 2^n$, with comparable computational cost and the same mean squared error loss function. The new work unifies these schemes and allows for the exploration of alternatives.

The main contribution is a variant that maintains the intuitive meaning of the scoring function but dispenses with auxiliary qubits and entangled measurements. This design results in a model more suitable for current NISQ devices, as it significantly reduces hardware demands without sacrificing the interpretability of the results. This optimization is crucial for the development of practical applications of quantum computing in structured information processing.