l2 if we made the lattice three dimensional we could

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 2. When x = 0, ⇡ ( 1) = 0 so p(0, 1) = 0 p(0, 1) = ✓/2 To check reversibility we note that if x p(0, 0) = 1 (✓/2). 0 then ✓ = ⇡ (x + 1)p(x + 1, x) 2 Here, as in most applications of the Metropolis-Hastings algorithm the choice of q is important. If ✓ is close to 1 then we would want to choose q (x, x + i) = 1/2L + 1 for L i L where L = O(1/(1 ✓)) to make the chain move around the state space faster while not having too many steps rejected. ⇡ (x)p(x, x + 1) = ✓x (1 ✓) · 38 CHAPTER 1. MARKOV CHAINS Example 1.36. Binomial distribution. Suppose ⇡ (x) is Binomial(N, ✓). In this case we can let q (x, y ) = 1/(N + 1) for all 0 x, y N . Since q is symmetric r(x, y ) = min{1, ⇡ (y )/⇡ (x)}. This is closely related to the method of rejection sampling, in which one generates independent random variables Ui uniform on {0, 1, . . . , N } and keep Ui with probability ⇡ (Ui )/⇡ ⇤ where ⇡ ⇤ = max0xn ⇡ (x). Example 1.37. Two dimensional Ising model. The Metropolis-Ha...
View Full Document

Ask a homework question - tutors are online