
- Sponsor
- Physics Department
- Speaker
- Tej Kanwar (University of Edinburgh)
- Contact
- Brandy Koebbe
- bkoebbe@illinois.edu
- Views
- 53
Lattice field theory simulation is a key non-perturbative approach to understanding strongly interacting field theories from first principles. Within particle and nuclear physics, lattice quantum chromodynamics (QCD) is now well-established as a leading method to obtain fully systematically controlled predictions of several low-energy QCD quantities, including hadron masses, parton distribution functions, nucleon form factors, and scattering/binding properties of multi-nucleon systems.
Extending the demonstrated success of lattice gauge theory into the upcoming years will however require further algorithmic improvements to overcome bottlenecks in the computational strategy, such as critical slowing down of Monte Carlo importance sampling and exponentially growing statistical noise. In this talk, I will discuss recently emerging ideas for applying machine learning methods towards this end. If the potential of these methods is fully realized, they stand to accelerate lattice gauge theory calculations by orders of magnitude while maintaining systematic exactness.