Current quantum state tomography (QST) algorithms are computationally expensive. This paper introduces Neural Adaptive Quantum Tomography (NAQT), a machine-learning-based algorithm that adapts measurements and significantly speeds up processing while maintaining high reconstruction accuracy. NAQT uses a neural network to replace the standard Bayes' update, enabling rapid integration of measurement adaptation and statistical inference. It's a meta-learning approach, learning to learn about quantum states without requiring an ansatz about the state's form. Despite its generality, it's retrainable within hours for two-qubit systems, suggesting feasibility for larger systems.

Yihui Quek, Stanislav Fort, Hui Khoon Ng