Berlin 2024 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

DY: Fachverband Dynamik und Statistische Physik

DY 5: Machine Learning in Dynamics and Statistical Physics I

DY 5.11: Talk

Monday, March 18, 2024, 12:15–12:30, BH-N 243

Tensor-network-based reinforcement learning for quantum many-body systems — •Giovanni Cemin — Max Planck Institute for the Physics of Complex Systems, Dresden, Germany

The exploration of quantum many-body systems is a widely pursued field. However, the exponential growth of the Hilbert space dimension makes it challenging to classically simulate quantum many-body systems and, consequently, to extract meaningful information. In this context, we present a novel framework for efficiently controlling quantum many-body systems based on reinforcement learning (RL). Our framework addresses the complexities of the quantum control problem by utilizing matrix product states to represent the many-body state within the trainable machine learning architecture of our RL agent. This novel methodology enables the control of systems on a scale beyond the capabilities of neural-network-only architectures, while retaining the advantages of deep learning, such as generalizability and robustness to noise. Notably, our research demonstrates the ability of RL agents to accurately handle previously unseen many-body states.

Keywords: Machine learning; Tensor networks; Many-body physics