Latency-aware adaptive optimizer for real-time frugal variational quantum algorithms

• Item type: SIG Technical Reports(1)
• 公開日: 2022-10-20
• タイトル: Latency-aware adaptive optimizer for real-time frugal variational quantum algorithms
• タイトル / 言語: en
• タイトル: Latency-aware adaptive optimizer for real-time frugal variational quantum algorithms
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Center for Quantum Information and Quantum Biology, International Advanced Research Institute, Osaka University
• 著者所属(英): en
• 著者所属(英): Center for Quantum Information and Quantum Biology, International Advanced Research Institute, Osaka University
• 著者名: Kosuke, Ito
• 著者名(英): Kosuke, Ito
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Variational quantum algorhtms (VQAs) are promising in near-term applications of quantum computers. For practical implementations of VQAs, efficient classical optimizers are actively being explored. Previous research proposed strategies to allocate the number of measurement shots per iteration of stochastic gradient descent by maximizing the gain per shot. Those strategies actually achieve fast convergence with respect to the total shot number. However, not only the shot number, we should take into account the circuit-switching latency as it is large compared to per-shot time in practice. In cloud services, per-task price is several orders of magnitude greater than per-shot price. In this work, we propose a strategy to adaptively allocate the appropriate shot number of each iteration to reduce the total real time or the cost for convergence by taking into account the latency time or per-task cost. We numerically show the efficacy of our algorithm.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Variational quantum algorhtms (VQAs) are promising in near-term applications of quantum computers. For practical implementations of VQAs, efficient classical optimizers are actively being explored. Previous research proposed strategies to allocate the number of measurement shots per iteration of stochastic gradient descent by maximizing the gain per shot. Those strategies actually achieve fast convergence with respect to the total shot number. However, not only the shot number, we should take into account the circuit-switching latency as it is large compared to per-shot time in practice. In cloud services, per-task price is several orders of magnitude greater than per-shot price. In this work, we propose a strategy to adaptively allocate the appropriate shot number of each iteration to reduce the total real time or the cost for convergence by taking into account the latency time or per-task cost. We numerically show the efficacy of our algorithm.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA12894105
• 書誌情報: 研究報告量子ソフトウェア（QS）; 巻 2022-QS-7, 号 27, p. 1-8, 発行日 2022-10-20
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2435-6492
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会