Speaker
nobuo sato
(Jefferson Lab)
Description
We present a new approach to performing Bayesian inference for QCD analysis of nucleon structure and hadronization, using machine learning to construct the inverse function mapping quantum correlation functions to observables. The new concept provides an alternative paradigm to the standard maximum likelihood or Bayesian posterior sampling methods. The effectiveness of the new technology is illustrated with application to the extraction of parton distribution functions from deep-inelastic scattering data, with results compared to recent global QCD analyses.
Primary author
nobuo sato
(Jefferson Lab)
