Summary
2023
Session Number:A2L-2
Session:
Number:A2L-22
Quantum-Walk-Based Bandit Algorithm and its Performance
Yamagami Tomoki, Segawa Etsuo, Mihana Takatomo, Röhm André, Horisaki Ryoichi, Naruse Makoto,
pp.26-29
Publication Date:2023-09-21
Online ISSN:2188-5079
DOI:10.34385/proc.76.A2L-22
PDF download (1.6MB)
Summary:
This study proposes a new scheme to solve multi-armed bandit (MAB) problems using quantum walks (QWs). QWs possess a unique property: the coexistence of linear spreading and localization. This property is utilized in various applications, but there are few examples of using it for decision-making. This paper realizes such an algorithm for MAB problems, one of the most fundamental decision-making models, by associating the two operations in a trade-off relationship: exploration and exploitation, with the coexisting behaviors of QWs. The study shows that this policy performs better than the corresponding random walk-based one.