Lienhard, Benjamin

Date:    Friday, Apr 8, 2022
Time:   11:00
Place:   ETH Zurich, Campus Hönggerberg, HIT F 11.1
Host:    Christopher Eichler

Machine Learning assisted Superconducting Qubit Readout

Benjamin Lienhard
Princeton University, USA

Quantum computers hold the promise to solve specific problems significantly faster than classical computers. However, the quantum processor’s constituent components, control, and readout must be very well-calibrated to realize a practical quantum computer. Over the last few decades, infrastructure and protocols have been developed to operate small-scale quantum processors efficiently. However, the operation of medium- to large-scale quantum processors presents new engineering challenges. Among those challenges are efficient and high-fidelity multi-qubit control and readout. In particular, qubit-state readout is a significant error source in contemporary superconducting quantum processors. I will discuss control and readout software tools for multiple superconducting qubits in this talk. We demonstrate deep machine learning techniques to improve frequency-multiplexed superconducting qubit readout pulse shapes and discrimination for a five-qubit system. Compared with currently employed readout methods, these novel techniques reduce the required measurement time, the readout resonator reset, and the discrimination error rate by about 20% each. The developed readout techniques are a significant step towards efficiently implementing near-term quantum algorithms based on iterative optimization and quantum error correction protocols necessary for future universal quantum processors.