Theoretical Physics Colloquium

Machine learning for lattice field theory

by Dr. Gurtej Kanwar (University of Bern, Switzerland)

Tuesday, January 25, 2022 from to (Asia/Kolkata)
at Zoom
Description
Lattice field theory provides a non-perturbative approach to understanding strongly interacting field theories, including, for example, the quantum chromodynamics (QCD) sector of the Standard Model. Among numerous other applications of lattice field theory, lattice QCD calculations will be crucial in the effort to pin down whether observed discrepancies with the Standard Model such as the muon anomalous magnetic moment and heavy meson decay rates are truly new physics effects. Unfortunately, state-of-the- art lattice calculations are limited by the enormous computational effort required for precision studies, motivating developments to lattice methods that reduce such costs. In this talk I will discuss recent progress in applying machine learning to lattice field theory in order to drastically improve efficiency. This approach introduces machine learning components in such a way that the exactness of the physics results is not compromised. We also find that it is important to incorporate certain symmetries exactly in the machine learning architectures used. These concepts are not limited to lattice field theory applications, and I will touch upon possible broader impacts of the ideas.