Realizing a reinforcement learning agent for real-time quantum feedback
| Quand ? |
Le 07/12/2022, de 14:00 à 16:00 |
|---|---|
| Où ? | Amphi H |
| S'adresser à | Christopher Eichler |
| Participants |
Christopher Eichler |
| Ajouter un événement au calendrier |
vCal
iCal
|
Rapid advancements in building quantum information processing devices at scale call for radical paradigm shifts in the real-time control of quantum systems. Reinforcement learning offers a powerful approach to learn sophisticated control strategies in the absence of detailed models and promises to become a game changer in quantum technology as it did in many other disciplines ranging from board games, to robotics, to fundamental science. In my talk, I will show how we achieve real-time feedback control of a quantum system by using a reinforcement learning agent. We realize the agent as a novel low-latency neural-network on a field-programmable gate array interacting with a superconducting quantum system at MHz rates, which is more than 100 times faster than any other previous implementation of a reinforcement learning agent deployed in any physics experiment. Our work paves the way towards using reinforcement learning for real-time control of quantum computers, most notably for quantum error correction and fault-tolerant gate operations. Beyond that, the unprecedented speed of the agent marks an important engineering achievement , which can be deployed in a wide range of other applications, such as real-time the control of optical systems.
K. Reuer, J. Landgraf, T. Fösel, J. O’Sullivan, L. Beltrán, A. Akin, G.J. Norris, A. Remm, J.-C. Besse, F. Marquardt, A. Wallraff and C. Eichler, arXiv:2210.16715 (2022)