Quantifying coherence and entanglement remains a major hurdle in the development of quantum technologies, especially when dealing with complex high-dimensional systems. Ting Lin, Zhihua Chen, and Kai Wu from Jimei University, along with Zhihua Guo from Shaanxi Normal University and Shao-Ming Fei and Zhihao Ma from Capital Normal University, demonstrated machine learning techniques to directly estimate these important quantum properties. Their work introduces a support vector regression (SVR) model that accurately determines coherence and entanglement using only minimal quantities that are easily measurable, specifically the diagonal elements and the trace of the density matrix. This innovative approach avoids the need for full-state tomography, greatly reducing experimental demands while maintaining accuracy. By employing support vector quantile regression, the team further ensures reliable and conservative estimates, paving the way for practical applications in quantum computing, sensing, and metrology.
Estimating quantum coherence and entanglement with minimal data
Scientists have made significant progress in efficiently estimating quantum coherence and entanglement, fundamental resources for emerging technologies. Their work presents a machine learning approach utilizing support vector regression (SVR) to directly estimate these quantum properties from minimal experimental data. The research team demonstrated that coherence estimation requires only the diagonal elements of the density matrix along with squared and cubed density matrix traces, while entanglement estimation requires additional squared and cubed density matrix traces. Experiments reveal that this method can significantly reduce resource overhead compared to full-state tomography while maintaining high accuracy in quantifying these quantum properties.
The researchers employed support vector quantile regression (SVQR) with pinball loss to prevent overestimation by the SVR model and ensured that more than 95% of the predictions were at a conservative lower bound. Remarkably, this lower bound reliability is maintained for over 93% of predictions even with 2% variation in input features, highlighting the robustness of our method. Data show that this approach leverages semidefinite programming (SDP) techniques to compute geometric measures of coherence and maximize fidelity between quantum states. The optimization process involves reformulating the problem for non-negative real variables. This allows the computation of geometric measures of coherence through SDP problems solved using the cvxopt solver in the picos library.
Additionally, the geometric measure of entanglement is computed using a lower bound derived from the maximum fidelity between the density matrix and the positive partial transpose (PPT) state. To train and validate the SVR model, scientists generated a dataset of 10,000 quantum states consisting of mixed, pure, and convex combinations of states of various quantum systems. This innovative technology provides a practical and scalable tool with potential applications across computing, sensing, and metrology, providing a route to characterize quantum resources with unprecedented efficiency.
Efficient quantum resource estimation using machine learning
In this study, we demonstrate a machine learning approach that leverages support vector regression (SVR) and support vector quantile regression (SVQR) to efficiently estimate quantum coherence and entanglement measures for unknown states. By using minimal experimental data, especially the diagonal elements of the density matrix and traces, the researchers were able to estimate resources such as the l1 norm of coherence, the relative entropy of coherence, and the geometric measure of entanglement. This means that the resource overhead is significantly reduced compared to traditional state tomography techniques while maintaining high accuracy.
The developed model consistently provides reliable lower bounds on the predicted values, with more than 93% of the estimates remaining safe even when the input data contains perturbations of up to 2%. This robustness is achieved through the implementation of SVQR with pinball loss, which effectively addresses the overestimation potential inherent in standard SVR models. Although this study acknowledges that there is variation in predictive performance between the 2-ctorit and 4×4 systems, the R2 values remain consistently high, indicating that reliability remains. Future work could extend this methodology to estimate other quantum correlations currently computed by positive semidefinite programming without requiring prior knowledge of the quantum states themselves.
