July 03 2017
05-07.07: Dr. Vedran Dunjko (MPI of Quantum Optics) visits us.
Dr. Vedran Dunjko (MPI of Quantum Optics) visits us from Wednesday, July 5th to Friday, July 7th. He will give a talk on Thursday, July 6th at 10:30 (room E.11, building E2.6).
"Advances in quantum reinforcement learning"
Abstract:
Quantum machine learning explores the interaction between quantum computing and machine learning, in both directions of influence. This emerging field has been trending in recent times for multiple reasons. First is due to promises that quantum computing can speed up big data analyses. The second stems from the converse direction of influence, as ML methods have been shown promising for the control quantum experiments by handling hard optimization tasks However, machine learning and artificial intelligence machinery can do more than analyze data and perform optimization. Indeed, progress in more general learning methods, such as reinforcement learning have driven some of the most exciting technological and scientific trends of recent times, such as the AlphaGo system. In this overview talk, we will present some of the results of the intersection between reinforcement learning and quantum computing. We will explain the basic ideas behind reinforcement learning on the example of Projective Simulation (PS), which is a physics-inspired learning model. Following this, we will review some of the main results exploring the mutually beneficial exchange between reinforcement learning and quantum computing including: quantum speed-ups of the PS model, generic quantum improvements in learning efficiency for reinforcement learning, but also most recent results showing how learning agents can be used to discover new quantum experiments. We will finish off with a brief discussion of what consequences these (predominantly theoretical) results may have on machine learning, artificial intelligence and quantum information processing in the near (and not so near) term.Some references:
-Projective simulation for artificial intelligence, H. J. Briegel & G. De las Cuevas Sci. Rep. 2, Article number: 400 (2012) -Projective simulation with generalization, A. A. Melnikov, A. Makmal, V. Dunjko, Hans J. Briegel, arXiv:1504.02247 - Meta-learning within Projective Simulation, A. Makmal, A. A. Melnikov, V. Dunjko, and H. J. Briegel, IEEE Access 4, 2110 (2016) [arXiv:1602.08017]. -Quantum speed-up for active learning agents, G. Paparo, V. Dunjko, A. Makmal, M. A. Martin-Delgado, and H. J. Briegel, Phys. Rev. X 4, 031002 (2014) [arXiv:1401.4997] - Quantum-Enhanced Machine Learning, V. Dunjko, J. M. Taylor, and H. J. Briegel, Phys. Rev. Lett. 117, 130501 (2016). -Active learning machine learns to create new quantum experiments, A. A. Melnikov, H. Poulsen Nautrup, M. Krenn, V. Dunjko, M. Tiersch, A. Zeilinger, H. J. Briegel
© Copyright 2010–2013 — Theoretical Quantum Physics Group
