Enhanced measurement of neutral-atom qubits with machine learning
Physical Review Applied (2024)
In this paper, scientists show that convolutional neural networks, or CNNs, can improve the accuracy of measuring the state of qubits in a neutral-atom quantum computer.
The researchers developed two types of CNNs: single-qubit and multiqubit. They compared the performance of these CNNs with a standard method. Their findings: Both types of CNNs significantly reduced errors in different scenarios, ranging from a 32% to a 56% reduction. The multiqubit CNN was especially effective at reducing the correlation between errors in neighboring qubits. It was able to distinguish between qubit states even when qubits were close together and potentially interfering with each other.
The paper is authored by L. Phuttitarn, B. M. Becker, R. Chinnarasu, T. M. Graham, and M. Saffman, who conducted the work at the University of Wisconsin–Madison and Infleqtio.