Conditioning categorical models to hard data using a Gibbs sampling of a truncated multi-variate Gaussian model

From a set of categorical model realizations that don’t necessarily match data, how do we generate categorical model realizations that fully honored data? This research proposes an efficient method to generate posterior categorical model realizations that match data 100% from prior categorical model realizations generated through any method. The proposed method is variogram-modeling free and relies on signed distance functions, principal component analysis, and Gibbs sampling. Francky Fouedijo