Quantum Algorithm Framework for Graph Bandit Problems Introduced in arXiv Preprint

A new quantum algorithmic framework known as Quantum Spatial Best-Arm Identification (QSBAI) has been developed to tackle best-arm identification challenges within graph bandit scenarios. This method enhances quantum reinforcement learning by incorporating spatial limitations dictated by graph connectivity. QSBAI employs quantum walks to represent superpositions over actions constrained by the graph, thus refining amplitude amplification strategies. It is based on the quantum BAI algorithm and utilizes Szegedy's walk framework. Theoretical evaluations have been conducted on complete and bipartite graphs, establishing upper limits for success probabilities. This research was published in the arXiv preprint arXiv:2509.05890v3, classified as a replace-cross announcement. Quantum reinforcement learning integrates quantum computation with sequential decision-making, with prior studies addressing the multi-armed bandit (MAB) problem. The graph bandit problem expands MAB by adding spatial constraints, and until now, quantum methods in this area have been scarce. The QSBAI framework aims to be relevant for a wide range of graph structures, rather than just specific types.