Mcmc is feasibe compared to iid sampling
If you have a high-dimensional probability density function, it’s difficult to sample independent and identically distributed values from from it. However, it’s feasible to use Markiv chain Monte Carlo sampling. MCMC sampling consists of a stochastic “walk” across the population space, where the next sample is determined by the previous one. After an extensive number of “steps,” the sampling density becomes a more and more accurate proxy for the probability landscape.