Efficient Readout Error Mitigation Using Singular Value Decomposition

• Item type: SIG Technical Reports(1)
• 公開日: 2021-06-24
• タイトル: Efficient Readout Error Mitigation Using Singular Value Decomposition
• タイトル / 言語: en
• タイトル: Efficient Readout Error Mitigation Using Singular Value Decomposition
• 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: Graduate School of Information Science and Technology, The University of Tokyo
• 著者所属: IBM Quantum, IBM Research Tokyo
• 著者所属(英): en
• 著者所属(英): Graduate School of Information Science and Technology, The University of Tokyo
• 著者所属(英): en
• 著者所属(英): IBM Quantum, IBM Research Tokyo
• 著者名: Bo, Yang; Rudy, Raymond
• 著者名(英): Bo, Yang; Rudy, Raymond
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Nowadays, the near-term quantum devices over 50 qubits are available and those of hundreds of qubits are expected to be realizable in several years. Mitigating their errors is one of the most important issues to exploit the better performance on such noisy devices. Readout error in the measurement process may have the biggest error rate among all errors, that can be mitigated by post-processing on classical computers. The standard readout error mitigation methods correct the measured probability distribution using the inverse of calibration stochastic matrices. However, such methods require computational resources exponential in the number of qubits, and are not applicable for devices with tens or more qubits. Here we investigate the polynomial time and memory classical post-processing based on the tensor product noise model. Our proposed algorithm requires O(ns2) time and O(s) memory with n qubits and s shots. In order to meet the physical constraint for the probability distribution, our algorithm uses Lagrangian multiplier on the singular value decomposition of calibration matrices, which analytically finds the closest solution. This idea might help us theoretically upper bound the closeness of mitigated vector to the ideal one. The current version of our proposed algorithm is targeted to mitigate the output probability distribution with few expected labels, which should be further explored and improved for the wider applications. To study the performance of the proposed algorithm, we report the result of numerical simulation on the modified Grover-search algorithms which are important for near-term applications of quantum devices. Our proposed algorithm shows almost the same performance as the conventional standard algorithms on 10-qubits system and is also likely to well mitigats the results on 20-qubits system where the standard algorithms would take prohibitively long time.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Nowadays, the near-term quantum devices over 50 qubits are available and those of hundreds of qubits are expected to be realizable in several years. Mitigating their errors is one of the most important issues to exploit the better performance on such noisy devices. Readout error in the measurement process may have the biggest error rate among all errors, that can be mitigated by post-processing on classical computers. The standard readout error mitigation methods correct the measured probability distribution using the inverse of calibration stochastic matrices. However, such methods require computational resources exponential in the number of qubits, and are not applicable for devices with tens or more qubits. Here we investigate the polynomial time and memory classical post-processing based on the tensor product noise model. Our proposed algorithm requires O(ns2) time and O(s) memory with n qubits and s shots. In order to meet the physical constraint for the probability distribution, our algorithm uses Lagrangian multiplier on the singular value decomposition of calibration matrices, which analytically finds the closest solution. This idea might help us theoretically upper bound the closeness of mitigated vector to the ideal one. The current version of our proposed algorithm is targeted to mitigate the output probability distribution with few expected labels, which should be further explored and improved for the wider applications. To study the performance of the proposed algorithm, we report the result of numerical simulation on the modified Grover-search algorithms which are important for near-term applications of quantum devices. Our proposed algorithm shows almost the same performance as the conventional standard algorithms on 10-qubits system and is also likely to well mitigats the results on 20-qubits system where the standard algorithms would take prohibitively long time.
• 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: 量子ソフトウェア（QS）; 巻 2021-QS-3, 号 12, p. 1-6, 発行日 2021-06-24
• 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