Machine Learning for Quantum Many-Body Physics
Coordinators: Roger Melko, Amnon Shashua, Miles Stoudenmire, and Matthias Troyer
Scientific Advisors: Juan Carrasquilla, Pankaj Mehta, Lei Wang, and Lenka Zdeborova
This KITP program will bring together experts from both physics and computer science to discuss the uses of machine learning in theoretical and experimental many-body physics. Machine learning will be explored as a complementary method to current computational techniques, including Monte Carlo and tensor networks, as well as a method to analyze "big data" generated in experiment. The program will include applications in the design of quantum computers and devices, such as the use of neural networks for the purposes of decoding, quantum error correction, and tomography. Theoretical connections between deep learning, the renormalization group, and tensor networks, will be examined in detail. Foundational questions in machine learning will be addressed, such as the formal concepts on information, intelligence, and interpretability. Finally, the theoretical possibility of a quantum advantage for machine learning applications implemented on near-term quantum hardware, such as quantum annealers, will be examined.
The program invites applications from researchers in condensed matter, quantum information, statistical physics, and related disciplines interested in exploring the interplay between quantum many-body physics and modern machine learning techniques; as well as computer scientists from the field of artificial intelligence research interested in sparking a dialog with physicists on these topics.