Reinforcement Learning to Manipulate Quantum Matter
Dr. Marin Bukov, University of Berkeley, California
Bibliothek, A.2.500, Staudtstr. 2
The ability to prepare a physical system in a desired quantum state is central to many areas of physics, such as nuclear magnetic resonance, quantum simulators, and quantum computing. Yet, preparing states quickly and with high fidelity remains a formidable challenge. I will introduce reinforcement (RL) learning ideas to manipulate quantum states of matter, and explain key practical advantages offered by RL. As a concrete example, I will demonstrate that RL allows to find short, high-fidelity driving protocols for transferring population from an initial to a target state in a non-integrable many-body quantum system of interacting qubits. Time permitting, I will show that quantum state manipulation, viewed as an optimization problem, exhibits critical phenomena, familiar from classical macroscopic systems: continuous and discontinuous phase transitions, symmetry breaking and spin-glass-like order occur in the optimization landscape and affect the learning behavior of the RL agent. I will highlight the potential usefulness of RL for applications in out-of-equilibrium quantum physics, and discuss potential future applications of RL to periodically-driven systems.