Robust quantum control in games: An adversarial learning approach

Ge, Xiaozhen; Ding, Haijin; Rabitz, Herschel; Wu, Re Bing

doi:10.1103/physreva.101.052317

Cited by 50 publications

(24 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…where represents the uncertainty parameter(s). Empirical analyses via stochastic gradient-descent algorithms [40,61] or min-max algorithms [62] were conducted to optimize robust control fields. The analyses collectively indicate that the algorithms almost always successfully find high-quality robust and high-precision control solutions.…”

Section: Unitary Control Objectivementioning

confidence: 99%

Optimization Landscape of Quantum Control Systems

Rabitz

2021

Complex Syst. Model. Simul.

Self Cite

View full text Add to dashboard Cite

Optimization is ubiquitous in the control of quantum dynamics in atomic, molecular, and optical systems.The ease or difficulty of finding control solutions, which is practically crucial for developing quantum technologies, is highly dependent on the geometry of the underlying optimization landscapes. In this review, we give an introduction to the basic concepts in the theory of quantum optimal control landscapes, and their trap-free critical topology under two fundamental assumptions. Furthermore, the effects of various factors on the search effort are discussed, including control constraints, singularities, saddles, noises, and non-topological features of the landscapes. Additionally, we review recent experimental advances in the control of molecular and spin systems. These results provide an overall understanding of the optimization complexity of quantum control dynamics, which may help to develop more efficient optimization algorithms for quantum control systems, and as a promising extension, the training processes in quantum machine learning.

show abstract

Section: Unitary Control Objectivementioning

confidence: 99%

Optimization Landscape of Quantum Control Systems

Rabitz

2021

Complex Syst. Model. Simul.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Such hybrid approach has been experimentally demonstrated with the nuclear magnetic resonance [131][132][133]. Other hybrid variants of GRAPE have also been proposed [134][135][136][137], which can be employed in quantum metrology.…”

Section: A Quantum Controlmentioning

confidence: 99%

Optimal Scheme for Quantum Metrology

Liu,

Zhang,

Chen

et al. 2021

Preprint

View full text Add to dashboard Cite

Quantum metrology can achieve far better precision than classical metrology, and is one of the most important applications of quantum technologies in the real world. To attain the highest precision promised by quantum metrology, all steps of the schemes need to be optimized, which include the state preparation, parametrization and measurement. Here we review the recent progresses on the optimization of these steps, which are essential for the identification and achievement of the ultimate precision limit in quantum metrology. We hope this provides a useful reference for the researchers in quantum metrology and related fields.

show abstract

“…The basic idea thereof is to formulate the robust control design as the minimization of an empirical loss function evaluated over uncertainty samples. The loss function is usually chosen as the average error or the worstcase error, based on which various gradient-based algorithms were proposed for the training of robust controls, e.g., the sampling-based learning [13,14], stochastic gradient-based algorithms [15,16], the sequential convex programming (SCP) [17,18] and the adversarial training based a-GRAPE [19]. These algorithms have been shown effective in im- * Electronic address: rbwu@tsinghua.edu.cn proving the robustness against various uncertainties, e.g., coupling uncertainty [15,19], energy broadening [13,14,20], inhomogeneity of control field pulses [13,14,21] and clock noises [22].…”

Section: Introductionmentioning

confidence: 99%

“…These facts show that the two loss functions make different trade-offs between the precision and the robustness. In our recent work [19], we illustrate that the a-GRAPE approaches can adjust the trade-off by purposely using poor-performance uncertainty samples. However, many hyper-parameters in the algorithm have to be empirically tuned, which is computationally costly when the control parameters or the uncertainty parameters are high-dimensional.…”

Section: Introductionmentioning

confidence: 99%

Risk-sensitive Optimization for Robust Quantum Controls

Ge,

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Highly accurate and robust control of quantum operations is vital for the realization of errorcorrectible quantum computation. In this paper, we show that the robustness of high-precision controls can be remarkably enhanced through sampling-based stochastic optimization of a risksensitive loss function. Following the stochastic gradient-descent direction of this loss function, the optimization is guided to penalize poor-performance uncertainty samples in a tunable manner. We propose two algorithms, which are termed as the risk-sensitive GRAPE and the adaptive risksensitive GRAPE. Their effectiveness is demonstrated by numerical simulations, which is shown to be able to achieve high control robustness while maintaining high fidelity.

show abstract

Robust quantum control in games: An adversarial learning approach

Cited by 50 publications

References 35 publications

Optimization Landscape of Quantum Control Systems

Optimization Landscape of Quantum Control Systems

Optimal Scheme for Quantum Metrology

Risk-sensitive Optimization for Robust Quantum Controls

Contact Info

Product

Resources

About