Freiburg 2019 – wissenschaftliches Programm
FM 23: Quantum & Information Science: Neural Networks, Machine Learning, and Artificial Intelligence I
FM 23.1: Invited Talk
Montag, 23. September 2019, 16:30–17:00, 3043
Learning to violate Bell inequality with reinforcement learning — •Alexey Melnikov, Pavel Sekatski, and Nicolas Sangouard — Department of Physics, University of Basel
Quantum experiments push the envelope of our understanding of fundamental concepts in quantum physics. The designing of modern quantum experiments is difficult and often clashes with human intuition. In my talk, I will address the question of whether a reinforcement learning agent can propose novel quantum experiments. In our works, we answer this question in the affirmative in the context of quantum optics experiments, although our techniques are more generally applicable. I will talk about reinforcement learning and demonstrate how the projective simulation model can be used to design quantum experiments and discover experimental techniques by considering two examples. In the first example, a reinforcement learning agent learns to create high-dimensional entangled multiphoton states . In the second example, our reinforcement learning agent learns to design quantum experiments in which photon pairs violate a Bell inequality. As a result of this learning process, the agent finds several optical setups with high CHSH values for various detection efficiencies, which is an important step towards realistic device-independent quantum cryptography. Our findings highlight the possibility that machine learning could have a significantly more creative role in future quantum experiments.
 A.A. Melnikov, H. Poulsen Nautrup, M. Krenn, V. Dunjko, M. Tiersch, A. Zeilinger, and H.J. Briegel. Proc. Natl. Acad. Sci. U.S.A., 115(6):1221, 2018