Exercise 14.21

Consider the query ${\textbf{P}}({Rain}{Sprinkler}{true},{WetGrass}{true})$ in Figure rain-clustering-figure(a) (page rain-clustering-figure) and how Gibbs sampling can answer it.

  1. How many states does the Markov chain have?

  2. Calculate the transition matrix ${\textbf{Q}}$ containing $q({\textbf{y}}$ $\rightarrow$ ${\textbf{y}}’)$ for all ${\textbf{y}}$, ${\textbf{y}}’$.

  3. What does ${\textbf{ Q}}^2$, the square of the transition matrix, represent?

  4. What about ${\textbf{Q}}^n$ as $n\to \infty$?

  5. Explain how to do probabilistic inference in Bayesian networks, assuming that ${\textbf{Q}}^n$ is available. Is this a practical way to do inference?

View Answer