Error reduction technique using covariant approximation and application to nucleon form factor

We demonstrate the new class of variance reduction techniques for hadron propagator and nucleon isovector form factor in the realistic lattice of $N_f=2+1$ domain-wall fermion. All-mode averaging (AMA) is one of the powerful tools to reduce the statistical noise effectively for wider varieties of observables compared to existing techniques such as low-mode averaging (LMA). We adopt this technique to hadron two-point functions and three-point functions, and compare with LMA and traditional source-shift method in the same ensembles. We observe AMA is much more cost effective in reducing statistical error for these observables.