Quantum state reconstruction for systems with different dimensions using machine learning

A machine-learning-based reconstruction system trained exclusively on m qubits is presented here to reconstruct a quantum state on systems of n qubits where m>n. This method eliminates the need to match the dimension of the system with the dimension of a model used for training. Additionally, we use the monotonicity property of the fidelity to relate the average reconstruction fidelity of m qubits to any lower-dimensional n qubits. We reconstruct the quantum states of randomly sampled one, two, and three qubit systems with a machine-learning model trained on four qubit systems. This approach provides a robust time-efficient machine-learning-based quantum state tomography as we reduce time required for training a model.