ABSTRACT
In the previous chapter, we introduced a noniterative sampling procedure (i.e., the exact IBF sampling) to obtain i.i.d. samples exactly from an observed posterior distribution for discrete MDPs. Generally, the exact IBF sampling is feasible only when the discrete latent vector (or missing data) Z is of low dimension. For more general cases (e.g., Problem 4.9, or when Z is discrete but of high dimension, or when Z is continuous and so on), the noniterative sampling method may need to be modified.
