Master thesis presentation Pontus Lindgren

Supervisors: Tahereh Abad and Anton Frisk Kockum
Examiner: Anton Frisk Kockum

Abstract: Better quantum gates are likely key to enabling fault-tolerant, useful quantum computers. This study has focused on comparing different approaches to single-qubit and two-qubit gate optimization. The primary focus is on deep reinforcement learning. It is shown that using an ansatz can enhance the performance of deep reinforcement learning, both for single-qubit and two-qubit gates, but only significantly for single-qubit gates. Two different approaches to including an ansatz are compared, and adding the ansatz to the output from the neural network might be preferred over pre-training. The study is unable to show a significant improvement of quantum gates using deep reinforcement learning relative to using a state-of-the-art black-box optimizer, despite the black-box optimizer being easier to implement experimentally. For any quantum gate defined by piece-wise constant controls, it seems likely that a low-pass filter can enhance the performance, at least if considered by the optimizer. This is significant as the bandwidth of experimental devices is limited, meaning that low-pass-like distortions might be expected in experiments. Finally, the study highlights the importance of ZZ-coupling to understanding two-qubit gates, a phenomenon not typically considered when analyzing two-qubit gates analytically.