smh0816 posted on 10-11-2010 7:13 AM
I heard that VMP is a generalization of Mean-field method,
which can be applied to MRF network similarly with Loopy belief propagation as introduced in Weiss’ 2001 paper:
“Comparing the mean field method and belief propagation for approximate inference in MRFs”
If I use VMP for a MRF network (with discrete unary and binary constraints),
is it the same with Mean field method introduced in Weiss’ 2001 paper?