Hello, everybody,
I am a newcomer here and fell into the markov chain
monte carlo area one month ago. Basically our research
is about inference algorithm using particle filter for
Bayesian network. we want to use Gibbs sampling or
metropolis-hasting method to move particles. The
problem puzzled me is, say, we know the full
conditional distribution of one component random
variable is the conditional distribution given its
markov blanket in the Bayesian model, which is
proportional to the product of the conditional
distribution of this variable given its parents and
the conditional distribution of each child of this
variable given their parents. But we don't know the
normalizing constant usually.
So, quesion is: in case we do not have an analytical
close form of the full condtional distribution, how
can we sample from the full conditional distribution
suppose we only know the full conditional distribution
is proportional to the product of some densities?
We are eager to have some hints or suggestions or
explanation of some gibbs sampling method to make our
research further. Any reply about the above question
will be highly appreciated.
Best wishes to all,
Wei Sun
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - Send 10MB messages!
http://promotions.yahoo.com/new_mail
------- End of Forwarded Message
