gaussian processes (GPs) are nonparametric priors over functions, and fitting a GP to the data implies computing the posterior distribution of the functions consistent with the observed data. Similarly, deep gaussian pr