If I want to compute the probability $p(y,z|\theta,\lambda)$,then how to? I know the answer is $p(y|z,\theta)p(z|\lambda)$, but I do not know how to? Please help me, thanks a lot
1 Answer
What you want to do compute is the joint probability of $p(y,z|\theta, \lambda)$. Therfore you need to multiply $p(y|z,\theta)$ and $p(z|\lambda)$.
If you look at the graph you can see that $y$ has $z$ and $\theta$ as parent nodes. This means that $y$ depends on the values of $z$ and $\theta$. In terms of probabilities this is expressed as $p(y|z, \theta)$. The vertical bar expresses conditional probabilities.
The same principle applies to $z$ which depends only on $\lambda$ and can therefore be expressed as $p(z|\lambda)$.
-
$\begingroup$ I see. $p(y,z|\theta,\lambda) = p(y|z,\theta,\lambda)p(z|\theta,\lambda)$ Since $y$ only depends on $z$ and $\theta$ and $z$ only depends on $\lambda$. Thus we obtain $p(y,z|\theta,\lambda) = p(y|z,\theta)p(z|\lambda)$. Am I right? I thought $\lambda$ influences $y$ through $z$. So I am not sure if $ p(y|z,\theta,\lambda)=p(y|z,\theta) $ $\endgroup$– DuFeiDec 10, 2015 at 12:10
self-study
question so please add the tag and tell us what you tried to solve the question, the background, and why it is important. $\endgroup$