Presentation Information
[10p-N202-6]Policy for Spatial Best-Arm Identification Problem via Quantum Walks
〇Tomoki Yamagami1,2, Etsuo Segawa3, Takatomo Mihana2, Andre Roehm2, Atsushi Uchida1, Ryoichi Horisaki2 (1.Saitama Univ., 2.Univ. Tokyo, 3.Yokohama Nat. Univ.)
Keywords:
quantum walk,multi-armed bandit problem,best-arm identification
In recent years, research on quantum reinforcement learning, which incorporates quantum computation into reinforcement learning, has been actively conducted, and its application to the multi-armed bandit (MAB) problem, or a fundamental issue in reinforcement learning, has also been reported. As an extension of the MAB problem, the graph bandit problem formulates decision-making under spatial constraints. However, no quantum approach has yet been proposed for this problem. In this study, we propose a best-arm identification algorithm for the graph bandit problem using quantum walks.