Spectral analysis by the method of consistent constraints

Two major challenges of numeric analytic continuation – restoring the spectral density, $s(\omega)$, from corresponding Matsubara correlator, $g(\tau)$ – are (i) producing the most smooth/featureless answer for $s(\omega)$, without compromising the error bars on $g(\tau)$, and (ii) quantifying possible deviations of the produced result from the actual answer. We introduce the method of consistent constraints that solves both problems.