
272 Current Trends in Bayesian Methodology with Applications
Ising and Potts models, the Swendsen-Wang Cuts (SWC) method [2] gener-
alized SW to arbitrary posterior probabilities defined on graph partitions.
The SWC method relies on a weighted adjacency graph G =< V, E >
where each edge weight q
e
, e =< i, j >∈ E encodes an estimate of the prob-
ability that the two end nodes i, j belong to the same partition label. The
idea of the SWC method is to construct a random graph in a similar manner
to the SW but based on the edge weights, then select one connected compo-
nent at random and accept a label flip of all nodes in that component with a
probability ...