Fast High-Fidelity Readout of a Single Trapped-Ion Qubit via Machine-Learning Methods

The accuracy and speed of qubit readout can greatly affect the performance of quantum computers, which are held back by the lack of a more adaptive, accurate method for determining the system's quantum state. This study uses field-programmable gate arrays for machine-learning-assisted methods of single-qubit readout on a Yb⁺ ion-trap system, achieving 99.53% average fidelity within 171 μs per sample. The proposed scheme shows considerable advantages over traditional methods in fidelity, speed, and robustness, and is compatible with real-time readout and feedback control of qubit states.