How does MAP (Maximum A Posteriori) estimation differ from MLE (Maximum Likelihood Estimation) through a Bayesian lens?