QHGRNN: A Quantum Hamiltonian Graph Recommendation Neural Network

Recommendation systems are required to discover reliable user-item preferences from sparse and high-dimensional interaction data, yet existing graph-based recommendation models often struggle to capture complex topological correlations, while quantum neural networks have not fully established an explicit connection between recommendation graph structures and quantum circuit construction. To address this issue, a quantum graph Hamiltonian learning formulation maps useritem graph information to the Hamiltonian of a topological quantum system and constructs parameterized quantum circuits according to Hamiltonian evolution. Based on this mechanism, a Quantum Hamiltonian Graph Recommendation Neural Network (QHGRNN) trains circuit parameters by minimizing the loss function with gradient descent. Experimental results on MovieLens-100K, MovieLens-Latest-Small, and FilmTrust using DeepQuantum and PyTorch show that graph encoding improves the test accuracy from 0.500 to 0.588 and reduces the test MSE from 0.250 to 0.231, while QHGRNN also exhibits competitive binary recommendation, rating prediction, Top-K recommendation, and noise robustness performance. Predictably, our model enriches quantum recommendation learning and provides a feasible route for future quantum computers to process highdimensional user-item graph data.