Adaptive filtering of projective quantum measurements using discrete stochastic methods

Abstract

Adaptive filtering is a powerful class of control theoretic concepts useful in extracting information from noisy data sets or performing forward prediction in time for a dynamic system. The broad utilization of the associated algorithms makes them attractive targets for similar problems in the quantum domain. To date, however, the construction of adaptive filters for quantum systems has typically been carried out in terms of stochastic differential equations for weak, continuous quantum measurements, as used in linear quantum systems such as optical cavities. Discretized measurement models are not as easily treated in this framework, but are frequently employed in quantum information systems leveraging projective measurements. This paper presents a detailed analysis of several technical innovations that enable classical filtering of discrete projective measurements, useful for adaptively learning system-dynamics, noise properties, or hardware performance variations in classically correlated measurement data from quantum devices. In previous work we studied a specific case of this framework, in which noise and calibration errors on qubit arrays could be efficiently characterized in space; here, we present a generalized analysis of filtering in quantum systems and demonstrate that the traditional convergence properties of nonlinear classical filtering hold using single-shot projective measurements. These results are important early demonstrations indicating that a range of concepts and techniques from classical nonlinear filtering theory may be applied to the characterization of quantum systems involving discretized projective measurements, paving the way for broader adoption of control theoretic techniques in quantum technology.

arXiv comment12 pages, 2 figures. (Appendices: 9 pages, 3 figures). Revised presentation; data, code-base and analysis unchanged

References in corpus (24)

• Supplementary information for "Quantum supremacy using a programmable superconducting processor"
• Past, Present, and Future of Simultaneous Localization And Mapping: Towards the Robust-Perception Age
• Manipulation and Detection of a Trapped Yb+ Ion Hyperfine Qubit
• Mitigating measurement errors in multi-qubit experiments
• Robust Online Hamiltonian Learning
• Adaptive Bayesian Quantum Tomography
• Quantum Kalman Filtering and the Heisenberg Limit in Atomic Magnetometry
• Robust Calibration of a Universal Single-Qubit Gate-Set via Robust Phase Estimation
• Multiparameter optimisation of a magneto-optical trap using deep learning
• Detecting and tracking drift in quantum information processors
• Experimental Adaptive Bayesian Tomography
• Efficiently measuring a quantum device using machine learning
• Single shot parameter estimation via continuous quantum measurement
• Practical adaptive quantum tomography
• Accelerated Randomized Benchmarking
• Experimental demonstration of cheap and accurate phase estimation
• Real-time calibration with spectator qubits
• Efficient estimation of resonant coupling between quantum systems
• Quantitative tomography for continuous variable quantum systems
• Quantum Model Averaging
• Integration of spectator qubits into quantum computer architectures for hardware tuneup and calibration
• Machine learning for predictive estimation of qubit dynamics subject to dephasing
• Autonomous adaptive noise characterization in quantum computers
• Structured Filtering

Cited by in corpus (2)

• A proposal of adaptive parameter tuning for robust stabilizing control of $N$--level quantum angular momentum systems
• Streaming quantum gate set tomography using the extended Kalman filter