Speakers
Daniel Hackett
Gurtej Kanwar
(University of Edinburgh)
Description
Normalizing flows provide a framework to learn statistically exact machine-learned maps between different lattice field theories. Flows constructed to map from QCD to the same theory with a (possibly localized) operator insertion provide a general tool to construct unbiased reduced-variance estimators for lattice QCD correlation functions. Building on previous applications to Feynman-Hellmann calculations, we extend this approach to include improved computations of two-point and three-point functions. We present preliminary results of several applications of this approach.
Authors
Daniel Hackett
Gurtej Kanwar
(University of Edinburgh)
