8813

In this question, we consider the application of importance sampling to Markov networks.

a. Explain intuitively why we cannot simply apply likelihood weighting to Markov networks.

b. Show how likelihood weighting can be applied to chordal Markov networks. Is this approach interesting? Explain.

c. Provide a technique by which the more general framework of importance sampling can be applied to Markov networks. Be sure to define both a reasonable proposal distribution and an algorithmic technique for computing the weights.