Start of funding 01.07.2023

Reinforcement learning adaptive execution of quantum gates

Dr. Markus Schmitt
University of Regensburg
Fakultät für Informatik & Data Science

Prof. Dr. K. Brigitta Whaley
University of California, Berkeley
Berkeley Quantum Information and Computaion Center



The realization of high fidelity quantum gates remains a central challenge for the productive use of quantum computers in the near future. The objective of this project is to find adaptive strategies for optimal quantum gate design. Using the framework of reinforcement learning, we will focus on two aspects, namely (i) the incorporation of “context” about device parameters and (ii) online control capable of responding to noise effects during the gate execution. (i) addresses the challenge that recalibrating a quantum chip means to optimize the pulse sequence for entangling gates between each possible pair of qubits on a chip. Our goal is to train a reinforcement learning agent to directly propose a suited control pulse just based on few characteristic physical parameters of the qubit pairs. The goal of (ii) is to open up a new approach to finding control strategies that not only counteract the effects of noise on average, but also exploit immediate feedback from partial observations of the systems to use optimal pulses for each individual gate execution. Both parts of this project will be building blocks towards the autonomous and efficient calibration of digital quantum computers.