Forums

Doubt in Belief Propagation

Started by maha devi August 1, 2010
Hi all,

I am currently working on Belief Propagation. I have a doubt while
estimating my beliefs. I need to estimate the r.v "xj's" of a vector "x"

i - function node
j -variable node

With some approximation, Message from function to variable node: Pi->j is
Gaussian and I have calculated its mean and variance.
Message from variable to function node: Pi<-j =P_x(xj) * Gaussian pdf (mean
and variance of this pdf known). Estimate of 'xj's' is the mean of the
distribution Pi<-j. I will then use this new estimate in next iteration and
repeat until convergence happens.

*I don't know how to calculate the mean and variance of Pi<-j =P_x(xj) *
Gaussian pdf . Here, **P_x(xj) = Gauss-Bernoulli Distribution*
*
*
Your reply would be of so much help!

Thanks a lot in advance

Regards,
Mahadevi