Speaker
Description
In lattice QCD studies, physical observables like the chiral
condensate or baryon number density are computed as the trace of a
combination of products of the inverse fermion matrix, typically
estimated stochastically using the random noise method. The accuracy of
this method depends on the number of random sources used; ideally, an
infinite number of sources would yield true physical results. However,
practical limitations introduce systematic errors due to the finite
number of sources. We propose using an unfolding algorithm based on a
sequential neural network to learn the inverse transformation from a
“true” distribution (obtained with a large number of random sources) to
a "measured" distribution (obtained with fewer sources). Applying this
learned transformation to observables measured with fewer sources can
improve accuracy. We demonstrate that this method's effectiveness is
strongly dependent on the distribution of the random noise vectors and
only weakly dependent on the matrix structure of the observable, making
it a viable approach for lattice studies.
