PPT Slide
Markov Random Fields (MRFs)(Besag, 1974, Journal of the Royal Statistical Society, Series B; also see Golden, 1996)
- MRF Definition
- Positivity Assumption
- Neighborhood Graph
- Constructing the MRF (Hammersley and Clifford, 1971)
p(x) = Z-1exp[-V(x)], where the MRF’s potential function V(x)=?Vc(?c) and ?c forms a clique of the Neighborhood GraphTheorem: Vc(?c) = 0 whenever ?c does not form a clique.
- Computing with the MRF (Besag, 1974; Geman & Geman, 1984)
- Sampling the MRF
- Expectations wrt MRF
- MRF Parameter Estimation
- MAP estimation