Characterizing the performance of quantum devices presents a critical challenge that requires a way to accurately capture both expected behavior and inherent uncertainties. Keio University's Poramet Pathumsoot, Michal Hajdušek, and Rodney Van Meter address this need with a new stochastic approach to Graybox characterization, a technique that combines known system dynamics with unknown transformations. Their research overcomes the important limitations of traditional greybox methods by incorporating uncertainty quantification, allowing researchers to make more informed decisions based on experimental data. The team demonstrates that models built using Bayesian neural networks accurately capture the distribution of observed data, surpassing the original Graybox method at up to 1.9time, providing flexible tools for device characterization even when detailed knowledge of noise models is lacking. This advancement promises to improve the reliability and interpretability of quantum experiments and pave the way for more robust quantum technology.
Quantum noise, machine learning, probabilistic programming
Research at the intersection of quantum computing and machine learning is rapidly advancing, driven by the need to overcome the challenges in building reliable quantum technologies. This work focuses on characterizing and mitigating noise in quantum systems. This is an important step in realizing the possibilities of quantum computation. Scientists are increasingly employing machine learning technologies, particularly Bayesian neural networks and stochastic programming, to improve quantum control, characterization, and optimization. These approaches allow researchers to model complex quantum systems and predict their behavior more accurately.
Key areas of the study include dynamic decoupling, comb-based spectral estimation, and characterization of non-Gaussian noise. Researchers are leveraging technologies such as Jax and Qiskit, along with uncertainty quantification packages such as UQPY to develop sophisticated models. This interdisciplinary field highlights the complexity and collaborative nature of this field of research, demonstrating a growing trend in applying machine learning to address fundamental problems in quantum computing.
Bayesian Grey Box Model Characterizes the Uncertainty of Quantum Devices
Scientists have developed stochastic greybox models to improve characterization of quantum devices, addressing the limitations of quantification of uncertainty. This new method combines known system dynamics with unknown transformations to exploit the strength of both “Whitebox” and “Blackbox” modeling approaches. By employing a single experimental procedure across multiple qubit realizations, the team trained a machine learning model to approximate unknown processes within the system and form a predictive graybox model. To explain the common stochastic noise in superconducting, nuclear spins, and kibits of trapped ions, the team implemented Bayesian neural networks in the “black box” component.
These networks use Bayesian inference to perform model parameters inference, and make predictions expressed as distributions rather than single values, and naturally quantify prediction uncertainty. Experiments that utilize binary measurement results streamline data analysis and increase efficiency. The results show that the stochastic gray box model surpasses the previous model at up to 1.9 times, significantly improving prediction accuracy and uncertainty quantification, providing a robust approach for local experimental development.
Probabilistic grey box model captures kit noise
Scientists have developed a stochastic greybox model to enhance characterization of quantum devices, addressing limitations of existing methods of quantifying prediction uncertainty. This work focuses on accurate modeling of quantum systems that undergo stochastic noise that leads to variation in device performance. Researchers analyzed the kit's data generation process and revealed that stochastic noise is distributed rather than a single value for the expected observable quantum. Experiments show that the newly developed stochastic greybox model outperforms the previous model by up to 1.
9time to capture observed data distributions. This improvement is attributed to the model's ability to accurately predict the expected value of shifted expectations. This is an important metric that is affected by stochastic noise. The team found that this ability to accurately capture expectations essentially limits the performance of predictive models. This study provides enhanced grey box characterization methods, provides better uncertainty estimation, and serves as a valuable tool for building reliable predictive models for quantum systems.
Stochastic grey boxes improve quantum gate calibration
This study successfully extends the method of characterization of grey boxes with robust uncertainty quantification, a key advance for reliable device analysis. The results show that this stochastic model significantly improves the capture of the observed data distribution, up to 1.9 times over the previous model. The team further demonstrated the practical application of the model by reconstructing the control calibration as a maximum likelihood estimation problem.
Applying the model to adjust the √x gate, a quantum gate, yielded control parameters that are closer to the globally optimal solution compared to their statistical counterparts. The analysis reveals that model performance depends on accurate predictions of expected values, highlighting key factors in characterization of effective quantum devices. The code developed for this research is published and will facilitate further investigation and application of this valuable tool for understanding and calibrating quantum devices.
