Quantum State Tomography with Machine Learning

Summary

An important challenge in building a quantum computer is quantifying the level of control obtained in the preparation of a quantum state. The state of a quantum device is characterized from experimental measurements, using a procedure known as tomography. Exact tomography requires a vast amount of computer resources, making it prohibitive for quantum devices larger than a few qubits. In this project we develop a practical, approximate tomography method using modern machine learning techniques. Our work is based on training artificial neural networks using measurement data obtained from a system of qubits. After training, the neural network is sampled to determine properties of the underlying quantum state. As part of a collaborative effort, we will demonstrate our machine learning algorithms on both synthetic and experimental measurement data. Our ultimate goal is to deliver practical machine learning technology to design and characterize near-term quantum devices.

Figure 1. A visualization of the phases of a quantum wavefunction of 100 qubits. At left, the exact value of the phases, obtained from a large-scale computer simulation. At right, the phases reconstructed with state tomography using artificial neural networks.