Reinforcement learning
Anna Dawid, Julian Arnold, Borja Requena, Alexander Gresch, Marcin Płodzień, Kaelan Donatella, Kim A. Nicoli, Paolo Stornati, Rouven Koch, Miriam Büttner, Robert Okuła, Gorka Muñoz-Gil, Rodrigo A. Vargas-Hernández, Alba Cervera-Lierta, Juan Carrasquilla, Vedran Dunjko, Marylou Gabrié, Patrick Huembeli, Evert van Nieuwenburg, Filippo Vicentini, Lei Wang, Sebastian J. Wetzel, Giuseppe Carleo, Eliška Greplová, Roman Krems, Florian Marquardt, Michał Tomza, Maciej Lewenstein, Alexandre Dauphin
Abstract
In this chapter, we introduce the field of reinforcement learning and some of its most prominent applications in quantum physics and computing. First, we provide an intuitive description of the main concepts, which we then formalize mathematically. We introduce some of the most widely used reinforcement learning algorithms. Starting with temporal-difference algorithms and Q-learning, followed by policy gradient methods and REINFORCE, and the interplay of both approaches in actor-critic algorithms. Furthermore, we introduce the projective simulation algorithm, which deviates from the aforementioned prototypical approaches and has multiple applications in the field of physics. Then, we showcase some prominent reinforcement learning applications, featuring some examples in games; quantum feedback control; quantum computing, error correction and information; and the design of quantum experiments. Finally, we discuss some potential applications and limitations of reinforcement learning in the field of quantum physics.