Speaker: Daniel Hackett (Fermilab)

Host: Srimoyee Sen

Title: Adventures in Krylov space: new methods for lattice QCD data analysis

Abstract: Lattice QCD computes the properties of hadrons by generating and analyzing noisy Monte Carlo estimates of hadronic correlation functions. This task is hindered by difficult-to-control systematics due to the compounding issues of exponentially degrading signal to noise and excited-state contamination. In this talk, I will motivate and discuss a new framework for the analysis of noisy correlation function data based on the idea of applying the Lanczos algorithm for eigendecomposition in the infinite-dimensional Hilbert space of the lattice theory. Versus previous approaches based on statistical fitting, the Lanczos framework offers rigorous bounds on systematic errors and other practical advantages. While explored here in the lattice context, the same ideas are more broadly applicable.