Direct Parameter Estimations from Machine Learning-Enhanced Quantum State Tomography

With the power to find best fit arbitrarily complicated symmetry, machine-learning (ML)-enhanced quantum state tomography (QST) has demonstrated its advantages in extracting complete information about states. Instead of using reconstruction model training a truncated density matrix, we develop high-performance, lightweight, and easy-to-install supervised characteristic by generating target parameters directly. Such model-based ML-QST can avoid problem dealing with large Hilbert space, but cab keep feature extractions high precision, capturing underlying symmetry data. experimentally measured data generated from balanced homodyne detectors, compare degradation noise squeezed states predicted models; both are agreement empirically fitting curves obtained covariance method. direct parameter estimations illustrates crucial diagnostic toolbox for applications states, process, metrology, advanced gravitational wave macroscopic generation.